Luminous Landscape Forum
Equipment & Techniques => Medium Format / Film / Digital Backs – and Large Sensor Photography => Topic started by: marc gerritsen on May 15, 2010, 07:18:38 am
-
you have a tilt shift camera and you fiddle about to get everything into the right plane on location or you don’t have one and you do it in post
either way your building is going to look very pointy on the top or your interior might have proportionally too much ceiling
in post though you can deal with both the perspective and also the disproportion of size
i have devised a way around that by using “progressive stretch”
how this works is that I divide the photo in increments
lets say i select the bottom 95% of the photo and stretch it by 5%
then the bottom 90% stretched again by 5%
i keep doing this until i reach the bottom
as you can see in the samples the the building which is progressively stretched isn't as "pointy" as the one which was just stretched
the windows progressively become smaller towards the top which ads to a more natural look
also the lobby shot which was progressively stretched, looks more natural as the top half and the bottom half are now in proportion
for those who think they don't want more work in post; it took 15 secs to straighten the photo vertically and a push of a button to set an action into motion for the progressive stretch.
for those who think that the photo would fall apart, think pixel interpolation; yes, you will end up with a lesser quality photo, but if you start off with a 110 mb tif file you can still end up with a 70mb tack sharp photo
the samples and actions are just a first try and need to be perfected so please critique the way of working but not the photo ....... as yet.
maybe i am barking up the wrong tree and there is software available that can do all this, but my extensive search has failed to come up with something
cheers
marc
-
you have a tilt shift camera and you fiddle about to get everything into the right plane on location or you don’t have one and you do it in post
either way your building is going to look very pointy on the top or your interior might have proportionally too much ceiling
You do not get the problem if you use movements... if you keep the sensor parallel to a rectangular building, it looks rectangular in the picture...
In your skyscraper shot, if the top of the building in the foreground comes opposite a point h% up the building behind (or the hth floor of a 100 floor building) that is how it will look (as a proportion of picture height, using shift to keep parallel) in the picture.
The unnatural look you get is the result of stretching the whole picture, and you (attempted to compensate) with progressive stretching.
Unfortunately, your technique makes square windows look rectangular in the upper floors.
It would be possible to correct the barrel distortion in the interior picture (bent ceiling line)... this would be a PITA, but it would increase the apparent height of the far corner of the room.
Does anyone know of any decent software to do this? Last time I tried it in PS you had to extend the picture so that the distortion was symmetrical (so that the fattest part of the barrel was in the middle of the picture) and in this case you would also have to split the picture in two.
-
Unfortunately, your technique makes square windows look rectangular in the upper floors.
It would be possible to correct the barrel distortion in the interior picture (bent ceiling line)... this would be a PITA, but it would increase the apparent height of the far corner of the room.
when you look at the unstretched photo the windows appear rectangular because they are further away from the viewer
i have done the same with the progressive stretch to create a more natural look
as far as the interior photo is concerned there is no barrel distortion as phocus certainly has taken care of it
so don't understand your comment there
-
when you look at this photo which was shot with a camera with movements
i find that the top of the building is way overstretched
and when you look at the different sections ticked in red
they are all the same size which is unnatural as the top sections should be smaller then the bottom sections
-
when you look at this photo which was shot with a camera with movements
i find that the top of the building is way overstretched
and when you look at the different sections ticked in red
they are all the same size which is unnatural as the top sections should be smaller then the bottom sections
Dear Marc,
To my eyes as an architect and arch. photographer, your progressive stretch simply destroys the proportions of the building..... it completely looks like the height of the floors are decreasing at the top - also, in the other image, the height of the round furniture looks quit tall, of course I have not seen them in real life - and again it's to my eyes
the last example shows, that this was maybe not the best viewpoint for this building.... I know - this could be the only spot were the camera could stand
very best,
adam
-
Dear Marc,
To my eyes as an architect and arch. photographer, your progressive stretch simply destroys the proportions of the building..... it completely looks like the height of the floors are decreasing at the top - also, in the other image, the height of the round furniture looks quit tall, of course I have not seen them in real life - and again it's to my eyes
the last example shows, that this was maybe not the best viewpoint for this building.... I know - this could be the only spot were the camera could stand
very best,
adam
Yes, I do have the same impression. When it comes to crucial details that's where the problems start.
Acheiving such perfection in PP is then a long task.
At the end, as I've asked before here, nothing seems today able to replace a proper view camera.
Maybe in a close future.
-
when you look at this photo which was shot with a camera with movements
i find that the top of the building is way overstretched
and when you look at the different sections ticked in red
they are all the same size which is unnatural as the top sections should be smaller then the bottom sections
The problem with this shot is that you're just too close; the FOV is too wide for a natural looking rectilinear perspective and there also could be some barrel distortion to make things worse. The best "fix" for this situation would be to back up a little, but of course that's no always possible. Sometimes when you're really close, tilting the camera up just enough to add a very small amount of keystoning will look more natural, because when we look at a building from that close in person we see some keystoning.
-
when you look at the unstretched photo the windows appear rectangular because they are further away from the viewer
i have done the same with the progressive stretch to create a more natural look
I am pretty sure that this effect you only get if you are correcting the perspective in PS and not optically. If you are parallel to a plane (or line) no matter how any 2 dimensional object (or 1 dimensional for a line) is moved about that plane, the object is not going to change shape. In the image I have below, the vertical perspective was corrected optically and all of the windows have the same proportion throughout the building (of course you need to compare windows in the same column since there is horiz distortion). Now I know there is horizontal perspective distortion but try to ignore that. All of the vertical lines on the windows (from the ones that are similar in real life) are equal throughout the image. I am sure if you find the area of each parallelogram formed from the windows in the image they will all be the same thus proving the proportions are equal.
For you first image, it just does not look right and I cant go into any further detail then that. Also, I am sure in the high res image the building appears less sharp at the top, something you would not get (or get as much since sharpness does falloff towards the edge of the circle) if you corrected optically.
For the last image you showed (which was corrected optically) that distortion is from the falling off effect you get from the edge of the image circle on wide lenses. If you could, stand further back and use a longer lens, or elevate yourself to a good height.
-
Actually Marc, I totally agree with you and it is an inherent issue when using movements. The light simply has further to travel to the edges of the film/chip plane and the image gets distorted. Buildings, no matter how tall, when shot close with a wide lens look extremely top heavy and it's been a peeve of mine for years. I've used similar solutions to yours at times in the past - this is more an art, than a science, right? You do what you think is right for your particular image. I don't find this so much an issue with interiors, but from time to time... All-in-all I agree with your thinking.
-
as far as the interior photo is concerned there is no barrel distortion as phocus certainly has taken care of it
so don't understand your comment there
Phocus is great, but to my eye, the ceiling line looks bent.
-
This is an odd discussion. Are you trying to capture the building as it really is? If you are using a view camera, this is the only way to do this accurately. Whether you think you see a buildings windows on the upper floor differently than the lower floors, it doesn't matter. In this example and in reality, the windows are the same size. Stretching them might feel good to you, but it's creating something that doesn't exist. Its just another way of making a building look like its falling backward.
-
While I find the decreasing window size in the shot of the building unsettling ( I prefer the "stretch the photo vertical" version). I've really got to applaud not just the fact that Marc is so creatively inventive, but that he is willing to share the results with us. That building could obviously not have been shot from that distance with a perspective control camera (or lens) while capturing the full height.
Here comes an interesting dilemma... The interpolation introduced in some applications of this technique might send some of us pixel peepers running for the hills, but if you're client is only ever going to use it on the web and in powerpoints AND the shot simply cannot be achieved with a plumb and level camera.... then get the friggin shot however you can and fix it!
For 99% of my work the image circles of my lenses accomplish what I need to do while keeping my camera quite parallel to mother earth, but 1% of the time....
Nice work, Marc.
-
thanks chris!
the building shot is indeed taken too close there was no way to go further back or go up
when i gave the photo to my client i did not do anything to it, i only used this building shot to demonstrate something,
as for as the building is concerned this might not be a good solution, must also agree with jonathan that the building
looks like it is leaning backwards a bit in the progressive stretch
as for as the interior shot i do think this is an improvement overall, as the bottom part and the top part are now more in proportion
i normally do not shoot interiors like that but in this case the focus is on the the glass sculpture hanging of the ceiling
so the only way is to shoot upwards (btw this is the largest glass sculpture in the world and runs the length of 84 m by nikolas weinstein)
and again there is no barrel distortion, if there is a perceived curving it is real, either the builders or the designers have done it but not the camera
i have tested phocus and flexcolor at nausea for this
anyway just wanted to share my rough findings of yesterday, i was really excited as i thought i was on to something,
i have never seen anybody do a progressive stretch
will certainly will do more refinement for the interiors but will leave the buildings alone
cheers
m
-
I agree with Chris!
Well done Marc.
Cheers,
Willem.
N.B Isn't there something in CS5 that does something like that?
-
there is software available that can do all this, but my extensive search has failed to come up with something
When looking for a software that could correct perspective by the numbers (of degrees of camera tilt), I found Digital Photo Shifter:
http://www.deraltenburger.de/Win-Tools/Index.html (http://www.deraltenburger.de/Win-Tools/Index.html)
It can do what I wanted, with or without proportional compressing of the upper part of high vertical objects. (The latter is what you want, if I got it right).
The tool comes in different versions:
1-as a standalone Windows app,
2-as a plug-in for PS
3-as a plug-in for FixFoto,
4-as a plug-in for the Windows version of PhotoLine
http://www.pl32.com/ (http://www.pl32.com/)
Only #4 offers 16 bit color depth. - German only...and I found the surface rather confusing.
Good light - Hening.
-
Here comes an interesting dilemma... The interpolation introduced in some applications of this technique might send some of us pixel peepers running for the hills, but if you're client is only ever going to use it on the web and in powerpoints AND the shot simply cannot be achieved with a plumb and level camera.... then get the friggin shot however you can and fix it!
Words of wisdom......
-
Words of wisdom......
yes but
this was an other test i have done to see exactly what happens with those interpolated pixels and realized that i can use this technique not just for power point and web
but print as well even used this shot on the front cover of a local developers magazine
A to E my 5-step perspective technique and G and H 100% comparison crops from before and after
when you look at the crops you will see that full page print would be easily achievable
what i do agree with 100% is to get the frigging shot anyway you can, make that client happy and get paid!!!
cheers
m
-
this is a better crop of the edge
-
Hi Marc,
I am interested in trying out your method, do you mind explaining it in more detail?
Thanks!
-
For new AP's......let me say also that sometimes......sometimes the best solution is NOT to correct the convergence and make the composition work without perspective correction. The cover of Antoine Predock's new Rizzoli monograph-Volume 5.
[attachment=22017:rizzoli5.jpg]
-
good shot and indeed agree 100% to not correct
m
-
When looking for a software that could correct perspective by the numbers (of degrees of camera tilt), I found Digital Photo Shifter:
http://www.deraltenburger.de/Win-Tools/Index.html (http://www.deraltenburger.de/Win-Tools/Index.html)
It can do what I wanted, with or without proportional compressing of the upper part of high vertical objects. (The latter is what you want, if I got it right).
The tool comes in different versions:
1-as a standalone Windows app,
2-as a plug-in for PS
3-as a plug-in for FixFoto,
4-as a plug-in for the Windows version of PhotoLine
http://www.pl32.com/ (http://www.pl32.com/)
Only #4 offers 16 bit color depth. - German only...and I found the surface rather confusing.
Good light - Hening.
will investigate, too bad only for pc so will try tomorrow at the office
viele dank
m
-
While I find the decreasing window size in the shot of the building unsettling ( I prefer the "stretch the photo vertical" version). I've really got to applaud not just the fact that Marc is so creatively inventive, but that he is willing to share the results with us. That building could obviously not have been shot from that distance with a perspective control camera (or lens) while capturing the full height.
Here comes an interesting dilemma... The interpolation introduced in some applications of this technique might send some of us pixel peepers running for the hills, but if you're client is only ever going to use it on the web and in powerpoints AND the shot simply cannot be achieved with a plumb and level camera.... then get the friggin shot however you can and fix it!
For 99% of my work the image circles of my lenses accomplish what I need to do while keeping my camera quite parallel to mother earth, but 1% of the time....
Nice work, Marc.
100+ degree lenses (like the Super-Angulon 47 XL and Apo-digitar 47XL) help, but, if you can shoot from the building opposite, (or a cherry picker, mast or whatever) my
[a href=\'index.php?showtopic=43338\']Virtual Viewpoint technique[/a]
can be very useful, and allow you to get the foreground in, and view the building from ground eye-level, without stretching.
-
100+ degree lenses (like the Super-Angulon 47 XL and Apo-digitar 47XL) help, but, if you can shoot from the building opposite, (or a cherry picker, mast or whatever) my
[a href=\'index.php?showtopic=43338\']Virtual Viewpoint technique[/a]
can be very useful, and allow you to get the foreground in, and view the building from ground eye-level, without stretching.
Maybe in theory. Has anyone other than yourself seen an example of this?
-
Maybe in theory. Has anyone other than yourself seen an example of this?
I have not done it myself yet, but I cannot see why it would not work.
-
100+ degree lenses (like the Super-Angulon 47 XL and Apo-digitar 47XL) help, but, if you can shoot from the building opposite, (or a cherry picker, mast or whatever) my
[a href=\'index.php?showtopic=43338\']Virtual Viewpoint technique[/a]
can be very useful, and allow you to get the foreground in, and view the building from ground eye-level, without stretching.
with all due respect, i really don't understand your theory
but it is probably me; i am not into theories, i always want to see examples as well
-
if you can shoot from the building opposite, (or a cherry picker, mast or whatever) my
[a href=\'index.php?showtopic=43338\']Virtual Viewpoint technique[/a]
can be very useful, and allow you to get the foreground in, and view the building from ground eye-level, without stretching.
I look really forward to see an image done with this technique.... really I think, it will be some kind of a challenge to mix, let's say 10 images, where everything in the background, and the building for that matter, have a different perspective, due to different hight of viewpoint of the time of capture.
I know from real life shooting, that if I move my viewpoint 10 millimeters when doing a stitch, for example with a dslr and a shiftlens, then nothing lines up perfect anymore....( thats why a tech camera, where the lens can be fixed when stitching, is so handy) so aligning 10 images, where you move your viewpoint 3 meters between each capture, is a thing I really look forward to see the outcome from.....
At best, I think you will see an image with 10 different perspective's - one for each viewpoint - how this will give you a feeling of viewing a building AND surroundings, from ground level is beyond me.... but please, prove me wrong with an example.
sorry for the rant, but you are presenting here an unproven technique as "very useful"
Marc here in the opposite, presents in a perfect way, a technique with a lot of samples and how to's - so everyone can enjoy it and use it if they want - thanks a lot Marc!
/adam
-
you are welcome adam!
you do some great work there in denmark!
also love the navigation of the projects on your website.
cheers
m
-
Marc, I think your approach works very well. I've had this problem more often with interiors than exteriors and hope you can be a little more precise in your explanation of the process. Perhaps a screen grab of the action would suffice. Thanks for posting your progress and while the interpolation is apparent, I agree with you that it's commercially acceptable. Jim
-
you are welcome adam!
you do some great work there in denmark!
also love the navigation of the projects on your website.
cheers
m
thank's a lot,
And it goes right back to you as well - your type of subjects is, IMO, often overdone with a lot of additional light on the scenes - your work seems to communicate the feeling of the space and object in a good way!
very best,
adam
-
I look really forward to see an image done with this technique.... really wink.gif I think, it will be some kind of a challenge to mix, let's say 10 images, where everything in the background, and the building for that matter, have a different perspective, due to different hight of viewpoint of the time of capture.
Yes, you will most definitely have major parallax issues with this approach. There's a guy who used this method to photograph redwoods, although instead of a cherry-picker or crane he gained height by climbing neighboring trees. The images are interesting because the tree is shown in correct perspective from a fairly close-up viewpoint, which makes them all the more impressive. The background actually repeats in layers though, because of the way perspective changes. It it's an interesting effect; and since the tree is the real subject, I don't really see it as a flaw in his photos. But somehow I doubt commercial architecture clients would find the result acceptable.
Here's a link (http://www.amazon.com/Tree-New-Vision-American-Forest/dp/1402767161/ref=sr_1_1?ie=UTF8&s=books&qid=1274066160&sr=8-1) to the book.
-
Marc, I think your approach works very well. I've had this problem more often with interiors than exteriors and hope you can be a little more precise in your explanation of the process. Perhaps a screen grab of the action would suffice. Thanks for posting your progress and while the interpolation is apparent, I agree with you that it's commercially acceptable. Jim
hi james
very happy to do that, but i have posted 3 samples and don't know which one you refer to
A the interior
B the building night shot
C the building day shot
cheers
m
-
I look really forward to see an image done with this technique.... really I think, it will be some kind of a challenge to mix, let's say 10 images, where everything in the background, and the building for that matter, have a different perspective, due to different hight of viewpoint of the time of capture.
Hi Adam,
This is not new, it is only restricted to aligning a single flat plane, e.g. facade, storefront, or mural, without any depth/foreground/background. As long as there is no depth in the image things can work just fine, PTAssembler even does it automatically with its Camera Position Optimizer (http://www.tawbaware.com/ptasmblr_help_camera_position.htm).
PTAssembler also offers adjustable projection methods that allow progressive compression towards edges/corners, much easier than the proposed stretching method (and with a better resampling method). As far as I know PTGUI also has a compression feature, but not as flexible as PTAssembler's choice of projections to mitigate the anamorphic distortion effect, including a hybrid method that can mix several of four projection methods (http://www.tawbaware.com/projections.htm) in 4 quadrants of the image.
BTW, the whole visual 'stretching' issue is caused by viewing the output image from the wrong position/distance (too far away), hence the anamorphic distortion. When the focal length is relatively short, we tend to view the image from too far away. When we view the rectilinear output image from a distance that corresponds to the focal length versus the sensor array size ratio, we would see the correct perspective.
Cheers,
Bart
-
I am more and more interested by Marc's approach I must say.
-
Marc, I think your approach works very well. I've had this problem more often with interiors than exteriors and hope you can be a little more precise in your explanation of the process. Perhaps a screen grab of the action would suffice. Thanks for posting your progress and while the interpolation is apparent, I agree with you that it's commercially acceptable. Jim
Marc, How about the first two examples. Thanks, Jim
-
Marc, How about the first two examples. Thanks, Jim
hi jim
i uploaded a rudementary iphone movie of a photo i worked on
chose a different photo then the other examples as this one can really demonstrate that the "progressive stretch" actually works
i realized i can not use it very much in architecture but for interiors it works wonders
again this is only to describe the concept and anyone can apply their own ways to it
cheers
marc
https://download.yousendit.com/bFFNbGtESEJreEJFQlE9PQ (https://download.yousendit.com/bFFNbGtESEJreEJFQlE9PQ)
-
Thank you Marc, I got it now! I appreciate you taking the time to illustrate the process. Jim
-
i am sorry, but i think this is not true.
as adamoerk already explained geometric proportions become completely false if this "progressive" stretch is applied. this does not look better , its just looking wrong and is not thought till the end.
of course windows at the top of a high building are more far away, and look smaller for this in "reality", but respecting this you should not apply any perspective correction. in any case you cant apply corrections to the vertical lines and not to the horizontals as well,- they have to be stretched if you want to maintain the same geometric forms in a building. and so function optical shift lenses as well ( they dont squeeze the image at the top "progressive" together, in fact they enlarge the top for the more angular entry of the light rays ). if you correct electronically you have to do simply the same.
if you want to respect the smaller scale of the upper end of the building you have to use the uncorrected image, because optically this is the correct one.
our brain is a clever animal, and so it corrects perspectives subjectively because we know that buildings stand straight up and so the building looks for us more natural in an image if it is again standing straight in an image, therefor we use shift lenses ( or correct it in ps ). but our bran does not "correct" vertical lines and horizontal lines not, quadrats still look as quadrats and dont transform to squares if they are more far away.
-
i am sorry, but i think this is complete bs. as adamoerk already explained geometric proportions become completely false if this "progressive" stretch is applied. this does not look better , its just looking wrong and is not thought till the end.
of course windows at the top of a high building are more far away, and look smaller for this in "reality", but respecting this you should not apply any perspective correction. in any case you cant apply corrections to the vertical lines and not to the horizontals as well,- they have to be stretched if you want to maintain the same geometric forms in a building. and so function optical shift lenses as well ( they dont squeeze the image at the top "progressive" together, in fact they enlarge the top for the more angular entry of the light rays ). if you correct electronically you have to do simply the same.
if you want to respect the smaller scale of the upper end of the building you have to use the uncorrected image, because optically this is the correct one.
our brain is a clever animal, and so it corrects perspectives subjectively because we know that buildings stand straight up and so the building looks for us more natural in an image if it is again standing straight in an image, therefor we use shift lenses ( or correct it in ps ). but our bran does not "correct" vertical lines and horizontal lines not, quadrats still look as quadrats and dont transform to squares if they are more far away. sorry that i cant find a better word than bs. for this theory. better you forget it.
sorry, but i don't like your tone!
i am just exploring different options and within an already busy schedule i take time out to share!
if you read the whole list of posts you will see that i have already agreed that this might not work very well for tall architectural buildings
and if you would have gone through all the posts then you might see that it actually works very well for interiors
if you would be so kind to show me with photos what you mean, i think that might be helpfull for all
cheers
m
-
sorry, but i don't like your tone!
i am just exploring different options and within an already busy schedule i take time out to share!
if you read the whole list of posts you will see that i have already agreed that this might not work very well for tall architectural buildings
and if you would have gone through all the posts then you might see that it actually works very well for interiors
if you would be so kind to show me with photos what you mean, i think that might be helpfull for all
cheers
m
you are certainly right marc, i wrote it with too fast mbp key pad .... so i unsharpened a bit my last comment because the post was not intended to harm anyone here and i appreciate the effort to try out things and to share them too ... thats always great.
in interiors it may happen that with wideangles you get stretched edges, esp. if there are objects closer to the camera as chairs at the side or - more so - round objects which arent centered. in that case i personally copy a layer of the selected zone and try to make it with the ps distort function looking more homogeneously, but i think this should be done motif depending and individually for each motif. there cant be a rule which you apply to all photography, because- interior or exterior - the interpretation of a static applied "progressive correction" results in wrong geometric renderings. there may be applications where this does not matter, but in architecture photography i believe it does.
-
Hi Marc,
I have done the "progressive stretching" or shrinking on quite a few wide angle shots in the past. Not AP shots but mainly shots involving people where you can't back up anymore and are deliberately shooting with a too wide lens to be able capture the entire scene. As everyone know people will be progressively stretched out as the a closer to frame.
After thinking long and hard about it I came to the conclusion that the filter > distort > spherize can do the same thing better and with less work.
The process for a AP shot like the ones you have above would be to:
1) do the perspective correction and "regular" stretching.
2) increase the size of the canvas to 200% with an empty area either on top or below but not both. You can do it interactively with the crop tool or by doing image > canvas size and then setting the height to 200% and clicking on one of the up or down arrows.
3) Go to filter > distort > spherize and set mode to "vertical only". Adjust the preview size and change the amount to taste.
4) Remove the extra canvas again with the crop tool or image > canvas size set to 50%.
Regards,
Peter
PS. I use CS2 so newer versions might have other options as well.
-
This is an effect that has always been there and I can remember at College being told it's sometimes better to under correct to keep it natural at the top. I also thought it unnatural to stand so close to a tall building and expect it to look straight. I assumed that a camera and lens worked like my eyeball does. Well it just went outside to catch some sun and was thinking about the points raised here. I looked at my house, I tilted my head up down and around, I expected to see the house change shape just like it would if I looked through a camera and moved it up and down. I've not done this before with my eye to look for changes, I was surprised to see no change in verticals or shape to the building, it was more like moving an enlarged image on screen around, I've never noticed that before. I did try it with a camera and the effect is completely different, I'm not sure why that should be, is it a brain or an optical design thing?
Kevin.
-
I did try it with a camera and the effect is completely different, I'm not sure why that should be, is it a brain or an optical design thing?
Hi Kevin,
A camera uses a so-called rectilinear projection. It projects the optical image on a flat plane. At extreme projection angles this will cause anamorphic distortion (think about text/numbers painted on a road). The distortion will not be noticeable when viewed from the correct position.
Cheers,
Bart
-
Hi Kevin,
A camera uses a so-called rectilinear projection. It projects the optical image on a flat plane. At extreme projection angles this will cause anamorphic distortion (think about text/numbers painted on a road). The distortion will not be noticeable when viewed from the correct position.
Cheers,
Bart
I assumed my eyeball worked the same way, obviously not. Strange after 40 years looking through a camera you think I would of noticed the difference.
Thanks,
Kevin.
-
Hi,
maybe the topic is already considered closed, but in case anybody is interested:
I am not sure what algorithm DXO uses but it might be worth a try to use it for architectural shots, if the elimination of anamorphic distortion is what you need. You can run a trial and import tiff, jpg or dng files afaik
http://www.dxo.com/us/photo/dxo_optics_pro...ns/anamorphosis (http://www.dxo.com/us/photo/dxo_optics_pro/optics_geometry_corrections/anamorphosis)
-
Hi,
maybe the topic is already considered closed, but in case anybody is interested:
I am not sure what algorithm DXO uses but it might be worth a try to use it for architectural shots, if the elimination of anamorphic distortion is what you need. You can run a trial and import tiff, jpg or dng files afaik
http://www.dxo.com/us/photo/dxo_optics_pro...ns/anamorphosis (http://www.dxo.com/us/photo/dxo_optics_pro/optics_geometry_corrections/anamorphosis)
Unfortunately, I think with serious AP we are talking about distortion in combination with movements (shift, rise, fall etc.) and DXO's lens correction profiles, via discussions I had with them last year, does not deal with either T/S lenses or shifted MF lenses.
-
I assumed my eyeball worked the same way, obviously not. Strange after 40 years looking through a camera you think I would of noticed the difference.
The image is not projected onto a flat surface in the case of the eyeball, the retina is concave.
That said, the keystoning effect does happen with human eye sight, just to a much lesser degree compared to a camera. You don't see the convergence when looking at your house, but if you stand at the base of 30-story building you'll definitely see it (but it still won't be as severe as when looking through a camera viewfinder with a wideangle lens mounted).
-
Unfortunately, I think with serious AP we are talking about distortion in combination with movements (shift, rise, fall etc.) and DXO's lens correction profiles, via discussions I had with them last year, does not deal with either T/S lenses or shifted MF lenses.
PTLens can do it. It's kind of klunky, and not 100% precise but it does work. They have PS actions that will expand the canvas to the size of the full image circle. Then you drag your photo to the correct position on the canvas and it does the correction.
-
hi jim
i uploaded a rudementary iphone movie of a photo i worked on
chose a different photo then the other examples as this one can really demonstrate that the "progressive stretch" actually works
i realized i can not use it very much in architecture but for interiors it works wonders
again this is only to describe the concept and anyone can apply their own ways to it
cheers
marc
https://download.yousendit.com/bFFNbGtESEJreEJFQlE9PQ (https://download.yousendit.com/bFFNbGtESEJreEJFQlE9PQ)
Thanks Marc for this great topic. This link is not working anymore, is the video available anywhere else?
-
I assumed my eyeball worked the same way, obviously not. Strange after 40 years looking through a camera you think I would of noticed the difference.
While it is true that the eye's cornea is comparable to a lens with the pupil acting as a diaphragm, the human eye works VERY differently in all other respects.
Only about 0.02% of the human retina is capable of acute resolvable vision. This area of the retina is called the macula lutea. You eye muscles basically move your eyeball rapidly across your field of view to focus portions of the object of interest onto this tiny area of the retina. Your brain then pieces this retinal information together to give you the perception of a single larger image. During the brain's reconstruction process, it accounts for the "optical distortion" in a very complex manner, unlike any camera photographic process.
If you had to make a comparison to a camera, I guess it would be closer to doing a huge amount of complex stitching with a very sophisticated tilt-shift lens.
-
An extremely interesting thread.
I am a relative amateur - although I earn modestly from time to time shooting (mostly) interior VR panos for websites. In addition I usually shoot some conventional solo shots, usually interiors, for my clients - frequently using the Nikon 14-24 mm 2.8 at its widest. So the issues discussed here are familiar to me, even though up to now I have only looked longingly at T/S lenses, much less view cameras. What I'd like to ask, since the contributors here show such a high level of expertise, concerns the Nikkor 14-24 although it obviously applies to other rectilinear wide-angle lenses. Perspective distortion aside, the peripheral stretching seen when using the wide end of this lens has always bothered me, although just how much it bothers me is determined by the nature of the subject matter.
At one point I downloaded an evaluation copy of the DxO software which offers volume anamorphosis correction:
http://www.dxo.com/us/photo/dxo_optics_pro...ns/anamorphosis (http://www.dxo.com/us/photo/dxo_optics_pro/optics_geometry_corrections/anamorphosis)
which seems to me to be an extremely useful feature. I have to confess that I have not purchased a copy of the application as in most other respects I'm quite happy using NX2 (bugs and eccentric UI notwithstanding.)
So my question is, how to effect the same correction of a single shot? I use PTGui quite a lot for both VR and other stitching projections and it seems possible that it may be useable for this purpose? I've managed to get reasonable results using "distortion torture" in PS but it feels very haphazard.
Apologies if this is a little off-topic. Another apology - I hadn't read the immediately preceding posts.
Roy
-
Marc, this is an interesting topic that deserves more analysis and discussion. I would like to see some examples of the same scene shot from the same position using both a wide-angle normal lens, and a less wide-angle shift lens, after the normal shot has been adjusted and corrected.
The 'free transform' controls in Photoshop allow for a variety of different types of perspective-type adjustments, including perspective, distort, and warp, which are the ones I use most.
What looks best is a matter of taste.
I recently returned from a major holiday in Europe and Russia where I took lots of photos of churches and various buildings at 14mm, using the Nikkor 14-24 on my D700. When attempting to correct what is clear to my eyes is an abnormal distortion, I encounter all the problems mentioned in this thread.
I simply don't agree that an image of any tall architectural structure, taken from a close distance (on the ground) with a wide angle lens, looks natural. Some correction is necessary.
Below is a shot I took in Cologne, of the Gross St Martin. I show the full, uncorrected image first, followed by the corrected image using free transform (perspective, distort and warp). I prefer the corrected version.
Does anyone prefer the uncorrected version? If so, why?
[attachment=23425:Great_St...d_6_0871.jpg] [attachment=23426:Great_St...d_6_0871.jpg]
-
I simply don't agree that an image of any tall architectural structure, taken from a close distance (on the ground) with a wide angle lens, looks natural. Some correction is necessary.
Exactly right. As I mentioned above, it can never look "natural" as the eye/brain is unable to reconcile a wide angle perspective. No amount of correction can fully account for this.
Below is a shot I took in Cologne, of the Gross St Martin. I show the full, uncorrected image first, followed by the corrected image using free transform (perspective, distort and warp). I prefer the corrected version.
The corrected version is less offensive to the eye, but it still has serious discrepancies making it difficult for the brain to reconcile the reality of the image. For example, the apparent size of the fountain in the foreground in relation to the buildings in the background at first suggests that the fountain must be approximately one story high and having the width close to that of one of the buildings. Yet, we know that it cannot be this large based on the proximity of the ground in the image and our experience with the size of actual buildings. In addition, the building on the right still looks like it is falling backwards, while the top of the building on the far left appears to be bending to the right. All of this is very disturbing to the eye, since it is not at all what we encounter when we simply look at things.
-
In addition, the building on the right still looks like it is falling backwards, while the top of the building on the far left appears to be bending to the right. All of this is very disturbing to the eye, since it is not at all what we encounter when we simply look at things.
The picture on the right still isn't properly corrected, this is probably why you think it looks like it is still falling backwards.
The waterwall looks larger because it is closer to the lens.
-
The waterwall looks larger because it is closer to the lens.
That's precisely my point of what makes the image look unnatural in comparison to one's visual perception. The near-far relationships of objects in a wide-angle photograph do not correlate with how the eye renders the same imagery directly. The more planar the content in the photograph, the less objectionable this becomes (in the sense of creating "natural looking" images). I do not mean to imply that wide angle perspective is in any way inferior, as it does provide a unique way of "seeing" things that our eyes cannot do on their own; and, of course, many photographers enjoy working with these lenses for this reason.
-
Does anyone prefer the uncorrected version? If so, why?
I prefer, between the two, the uncorrected version, the reason being that in the corrected version the tower looks much "fatter" than it is in the original and, I do presume, in reality. Trying to correct keystoning in software in my limited experience stretches and deformates objects by making them lower and fatter.
So after having done what you have done to the image, one might apply a second-pass of deformation by re-stretching again the image so that the tower looks again having the same height, and the same height-width proportions, as in the original image. But by then one will have applied two rounds of stretching, quality will certainly have suffered somehow, and the final result might look not so good anyway.
Generally speaking I tend to try to correct perspective only when there is a tall element (such as this tower) at one side of the image, clumsily converging toward the "empty" centre of the image. When the tall element is in the center of the picture I often find that even with a 24mm equivalent the converging lines are good-looking and overall don't give an unnatural rendering of the object matter. Wide-angle lenses can generate images which are up to a certain extent unnatural, but the more wide angle become widespread the more the casual observer will get a trained eye for that and will consider the "distortion" as a natural characteristic of the medium, a bit like black-and-white pictures, that don't exist in reality but are not perceived as "unnatural".
As a final consideration I would add that even when using a PC lens I normally don't correct keystoning completely, as IMO leaving some convergence of lines makes the picture more natural, while making parallel those lines that, in the reality of human vision, are not parallel would make the picture, in general, slightly unnatural.
But overall I tend to leave a lot of this kind of images totally "uncorrected".
Cheers
Fabrizio
-
As a final consideration I would add that even when using a PC lens I normally don't correct keystoning completely, as IMO leaving some convergence of lines makes the picture more natural, while making parallel those lines that, in the reality of human vision, are not parallel would make the picture, in general, slightly unnatural.
But overall I tend to leave a lot of this kind of images totally "uncorrected".
Cheers
Fabrizio
Are you saying that a building doesn't have straight sides? When I look at a building I see the sides as parallel to each other. Even if my eyes showed convergence, in reality we know that the sides are parallel. It's the job of a photographer to capture reality - or at least it's the architectural photographer's job.
-
I prefer, between the two, the uncorrected version, the reason being that in the corrected version the tower looks much "fatter" than it is in the original and, I do presume, in reality. Trying to correct keystoning in software in my limited experience stretches and deformates objects by making them lower and fatter.
So after having done what you have done to the image, one might apply a second-pass of deformation by re-stretching again the image so that the tower looks again having the same height, and the same height-width proportions, as in the original image. But by then one will have applied two rounds of stretching, quality will certainly have suffered somehow, and the final result might look not so good anyway.
Generally speaking I tend to try to correct perspective only when there is a tall element (such as this tower) at one side of the image, clumsily converging toward the "empty" centre of the image. When the tall element is in the center of the picture I often find that even with a 24mm equivalent the converging lines are good-looking and overall don't give an unnatural rendering of the object matter. Wide-angle lenses can generate images which are up to a certain extent unnatural, but the more wide angle become widespread the more the casual observer will get a trained eye for that and will consider the "distortion" as a natural characteristic of the medium, a bit like black-and-white pictures, that don't exist in reality but are not perceived as "unnatural".
As a final consideration I would add that even when using a PC lens I normally don't correct keystoning completely, as IMO leaving some convergence of lines makes the picture more natural, while making parallel those lines that, in the reality of human vision, are not parallel would make the picture, in general, slightly unnatural.
But overall I tend to leave a lot of this kind of images totally "uncorrected".
Cheers
Fabrizio
The same effect will happen with a view camera if using a wide angle w/ movements, and there are objects nearby.
I used to have a book on scheimpflug published by Sinar that had a tilt degrees point past which you should consider *not* using movements (or using less, and tilting up some) because the result may be unnatural. Human vision does start to keystone after a certain tilt of the head. I wish I could remember the exact amount, but usually I notice it in the viewfinder anyway.
-
Exactly right. As I mentioned above, it can never look "natural" as the eye/brain is unable to reconcile a wide angle perspective. No amount of correction can fully account for this.
The corrected version is less offensive to the eye, but it still has serious discrepancies making it difficult for the brain to reconcile the reality of the image. For example, the apparent size of the fountain in the foreground in relation to the buildings in the background at first suggests that the fountain must be approximately one story high and having the width close to that of one of the buildings. Yet, we know that it cannot be this large based on the proximity of the ground in the image and our experience with the size of actual buildings. In addition, the building on the right still looks like it is falling backwards, while the top of the building on the far left appears to be bending to the right. All of this is very disturbing to the eye, since it is not at all what we encounter when we simply look at things.
......the building on the right still looks like it is falling backwards, while the top of the building on the far left appears to be bending to the right.
Quite right. I agree. And that's what disturbs me about the image. It needs more work in this regard.
However, I'm puzzled by your following comment.
...the apparent size of the fountain in the foreground in relation to the buildings in the background at first suggests that the fountain must be approximately one story high and having the width close to that of one of the buildings. Yet, we know that it cannot be this large based on the proximity of the ground in the image and our experience with the size of actual buildings.
I don't see this size issue as a disturbing problem. The brain is well-used to assessing distances by the relative size of objects. When a large building appears small in relation to a small object that appears big, we assume that we are very close to the small object, and relatively far from the big object. So it is the case with the fountain and statues. The statues were one of the main features in the composition, if not the main feature, and I was very close indeed when I took the shot using a 14mm lens.
There is nothing in front of the fountain to cause confusion about its size and everything that is behind the fountain looks, to my eyes, appropriately diminished in size (approximately), indicating its greater distance from the viewer.
However, I agree there is something disturbing about the height (or perspective) of the fountain and that I believe is the result of an impression created by the parallel walls of the buildings, rather than their size. There's a sense that the viewer is about 6 metres high (that is, the shot was taken from across the street from a window about 6 metres off the ground, in another building). Such an impression might be in conflict with the perspective one gets of the two statues on the fountain. If the shot was taken from a 6 metre height, then the statues must be very high off the ground because we are not getting a perspective of the top of their heads in the photo.
What I'd like to know is, if this effect can be avoided using a shift lens. I did consider buying the new Canon 17mm TSE for this latest trip, but decided against it on weight grounds. The D700 with 14-24/2.8 is heavy enough. The Canon 50D with EF-S 17-55/2.8 was a lighter option as a second camera than a 5D with 17mm TSE plus a 24-105/F4 zoom. I might have made a mistake, though.
-
One fascinating aspect of this issue in this thread, of how best to create a natural perspective in archetecture, is the other issue of the true focal length of the average human eye.
If one does a Google search on this question, one gets a variety of answers ranging from 17mm to 50mm. What the heck's going on here?
So I picked up my D700 with 14-24 attached, and 50D with 17-55 attached, and stood in the corner of my lounge, about 1.5 metres from my laptop, chair and small table.
The chair was the closest to the camera, and the front door, diagonally at the opposite corner of the room about 8 metres away.
I stared at the door, then raised the camera to my eye, several times at various focal legths using both cameras.
My impressions are, at 14mm (full frame 35mm) the view through the finder appears to approximate to my eye's angle of view including the extreme peripheral vision which has totally lousy resolution and which is only good for movement detection. In fact the peripheral vision through two eyes might be slightly greater than a 14mm angle of view, full frame DSLR, but I have no doubt which photo would be sharper towards the edges, comparing the D700 at 14mm with a snapshot of my brain's impression.
What's also very obvious is that, at 14mm, my front door appears much smaller through the viewfinder than it does when not looking through the camera. The effect of 'smallness' of an object just 8 metres away continues at higher focal lengths until we reach about 45 to 50mm.
At 50mm, when I toggle between gazing through the viewfinder and staring straight at the door, I see no noticeable change in the size of the door. However, a 50mm angle of view crops off most of that peripheral vision.
I get the impression (just a rough impression) if I were able to take a recordable snapshot of my brain's impression of the view of the front door, then crop off all the peripheral fuzzy stuff (and some of that peripheral stuff is a lot fuzzier than the out-of-focus areas of a shot with a 50mm lens at F1, including the worst corner and edge degradation of any lens every made), then I would get a fairly decent analogy with a photo taken with a 50mm lens.
The attraction of wide-angle lenses is that the camera is able to record in fine detail scenes which the human eye cannot encompass from one, fixed perspective. The eye can of course move around and take in a very wide scene in all its detail, but not in one fixed stare, even if the eyeballs are moved.
That's why I find the following shot interesting, taken recently in Vienna. There's no perspective correction here, by the way. This is straight out of the camera, uncropped. D700 with 14mm at F16. (Sharp from big toe to infinity ).
[attachment=23430:DSC_1387.jpg]
-
Are you saying that a building doesn't have straight sides? When I look at a building I see the sides as parallel to each other. Even if my eyes showed convergence, in reality we know that the sides are parallel. It's the job of a photographer to capture reality - or at least it's the architectural photographer's job.
Jonathan, what is reality? Are rails of a railway "parallel" or "converging"? They are parallel but, for a well-known law of nature, they appear to be converging. In reality they are (visually) converging and not parallel. Axonometry does not exist in nature, all that lives sees the world according to the laws of perspective.
Railways are not different from corridors, and corridors are not different from skyscrapers seen from below. If we had the habit of walking upward or if we could fly, it would seem to us as meaningless to try to make building sides "parallel" as it would seem absurd to try to make corridors walls "parallel". If swallows took pictures of buildings, they would never use PC lenses...
Do this experiment: put yourself in front of a convex corner of inside your house, at a distance of let's say 1,5 metres. Now tilt your head up and down with movements of around half a second each (in one second you will have gone from floor to ceiling to floor again). You will clearly see that your eyes record the vertical line between floor and ceiling as being curve (it will come "toward" you at its nearest point to your eyes).
Now, guess what: that line IS curve to your eyes and cannot be otherwise. You know, and I know, that from an "engineeristic" point of view that is a straight line. But from a visual point of view, that cannot be a straight line because where the line intersects the floor, and the ceiling, are points which are further than the nearest distance between you and the wall.
Now repeat it with any other element. Put yourself under a beam of your house. Look upward, and swing your head along the beam. See how curved the beam is? To your eyes, it is curved. Physically it is not. Visually it certainly is.
Now you might say: OK when I am near the corner, and look fast upward and backward, the effect is quite apparent but in my normal experience this effect is so small and my brain automatically makes me not to notice it.
Well, yes and no. I can be very easily aware that the sides of the inside of any builiding are NOT parallel when I look at anything. I can go round my house and be perfectly aware that my eyes look at it as if I were in, so to speak, a spheric environment. We all look at the world like a toad does, only with a less extreme perspective. It is a difference of degree, not of nature.
IMO our problem is how to make a certain picture look "natural". If you take pictures of a tall building, looking up, parallel lines will not necessarily look "natural". If you repeat your experiment with outside buildings and if they are tall enough that the perspective is evident, you will clearly see that the buildings don't appear to have parallel lines at all if you go enough under them and look at them upward.
Up to a certain extent our brain "stretches" buildings sides, but this "mental" effect is always only partial and up to a certain extent our mind is at the same time always aware of the perspective surrounding us.
Cheers
Fabrizio
-
Fabrizio,
I understand your point. You must already know that the perspective shown with railroad tracks and a corridor are very different than the convergence of vertical lines. You will not be able to convince an architect or an architectural photographer that those vertical lines should be left alone. It's our job to capture the building as it is designed. If you think a photograph looks better uncorrected than that's fine. I guess it's all a matter of opinion.
-
There is nothing in front of the fountain to cause confusion about its size and everything that is behind the fountain looks, to my eyes, appropriately diminished in size (approximately), indicating its greater distance from the viewer.
The perspective in your image is not "true perspective", as the relative scales of the near-far objects are not natural in the image (again, where natural refers to what the human eye perceives). The chairs, tables and even doors of the central buildings are much too small to account for the ground distance, while the opposite is true for the fountain in the foreground. Shooting the same scene with a larger format camera and an appropriately longer focal length lens would allow you to see this better. Alternatively, you can try to simulate such perspective by digitally combining smaller format images. For example, if you use a normal focal length lens and stitch images together to form the same field of view (and most likely do some focus blending as well), you would then be able to restore a more natural balance of scale.
-
The perspective in your image is not "true perspective", as the relative scales of the near-far objects are not natural in the image (again, where natural refers to what the human eye perceives). The chairs, tables and even doors of the central buildings are much too small to account for the ground distance, while the opposite is true for the fountain in the foreground. Shooting the same scene with a larger format camera and an appropriately longer focal length lens would allow you to see this better. Alternatively, you can try to simulate such perspective by digitally combining smaller format images. For example, if you use a normal focal length lens and stitch images together to form the same field of view (and most likely do some focus blending as well), you would then be able to restore a more natural balance of scale.
I would certainly agree that my shot is not a true perspective as seen by the eye from the position I took the shot. It cannot be. I used a 14mm lens from a distance of about 1 to 2 metres from the statue (can't remember the precise distance, but I was very close). From that distance, the surrounding houses to the left and right would have been an unidentifiable blur. If I were to change the direction of my gaze, to the left or the right, so that I were able to see the houses clearly, the statues in front of me would become an unidentifiable blur. Hence my reference in an earlier post to the focal length of the human eye, which seems to be another complicated issue. I get the impression that the human eye is a bit like looking through a 50mm lens on a large format camera, which has atrociously bad edge resolution. So bad that the resolution begins to fall off just a short distance from the centre.
I don't agree with those who claim that the image looks more natural when uncorrected for perspective. It's true that parallel lines converge in the distance, and large objects appear smaller in the distance, and that's information the brain uses to guauge distance. However, buidings to the side of one do not appear as though they are about to topple because they are leaning so steeply. If I were actually able to see that brick building on the far right of the uncorrected photo, leaning to the degree it is, as I stood in front of the fountain, I would get very alarmed.
The problem as I see it, is how to correct the perspective so that the scene looks as though the shot were taken from a different perspective which more closely corresponds to the eye's rather narrow FoV. This other perspective must be from a greater distance. In my perspective-corrected shot of the St Martin's church, it's unlikely that the photographer was positioned 6 metres off the ground across the street, because that would mean that the fountain with statues would be unusually tall (not an impossibility, but unlikely). If the statues weren't unusually tall, then from that quite close position across the street, I would get a view of the top of the statues' heads, which I don't.
Therefore, in my opinion, as I try to make perspective corrections in this photo, I'm trying to create the impression that the shot was taken from a much greater distance than it actually was, using a much longer focal length of lens, perhaps a 200mm lens from a distance of 100 metres from the fountain, whatever.
So the question I would ask is, have I succeeded in creating this false impression, that I took the shot from a distance of 50-100 metres away using a fairly long focal length of lens? (on the basis of course that one doesn't have reference to the uncorrected image which gives the game away).
-
The perspective in your image is not "true perspective", as the relative scales of the near-far objects are not natural in the image (again, where natural refers to what the human eye perceives). The chairs, tables and even doors of the central buildings are much too small to account for the ground distance, while the opposite is true for the fountain in the foreground. Shooting the same scene with a larger format camera and an appropriately longer focal length lens would allow you to see this better. Alternatively, you can try to simulate such perspective by digitally combining smaller format images. For example, if you use a normal focal length lens and stitch images together to form the same field of view (and most likely do some focus blending as well), you would then be able to restore a more natural balance of scale.
Sorry, it just doesn't work this way. You would need to change the postion of the camera to change the perspective. Changing focal length or sensor/film size doesn't change perspective. (See multiple threads discussing this on this and nearly every other photographic forum)
-
Since I'm in a creative mood on a high intellectual level (after a couple of glasses of wine ) I'd like to elaborate on this perspective issue. I once made a remark on this site a couple of years ago, to the effect that we all know that wide-angle lenses can create an impression of a different perspective. All hell broke loose and 100 posters tried to convince me that I was dead wrong and that focal length had nothing to do with perspective, and that perspective was affected only by the distance to the subject.
Their empirical evidence for such a declaration was based upon the fact (with which I agree) if one crops a shot taken with a wide-angle lens to the same FoV as any shot taken from the same position and direction with any other focal length of lens, then the size relationships between obects in the foreground and objects in the background will be the same.
Of course, the claimants of this theory (Jack Flesher and Gabor spring to mind) seemed oblivious the fact that the choice of FL of lens determines in practice the 'impression' of the distance from the viewer (or photographer) to the subject.
For example, if I take a shot of an isolated figure on a hill top, who is immediately in front of a huge mountain in the background, it makes no difference to the relative sizes of the figure and the mountain, whether I use a normal lens or a telephoto lens. That's true and cannot be denied (discounting lens distortions, of course).
But what some people fail to realise, is that the choice of focal length may affect the viewer's impression of the distance to the nearest subject in the composition. The choice of focal length doesn't affect the spatial relationships between the nearest point and the furtherst point in any particular crop, but it does affect the sense of distance to the nearest point (in the composition) to the viewer.
Okay! Forgive my rant. You claim the chairs against the building on the left appear too small and that, therefore, the statues are unnaturally large in comparison.
In my defense, I'm posting crops of the statues and chairs from both the unadjusted and adjusted images previously posted. Now I agree that the first impression is that my perspective correction has made those two women appear too squat. That's a concern which I may correct, but I'll say that those two women caught my attention because they were such squat little dumplings, one of them obviously bemoaning the fact that they had yet another church after that grotesque monstrosity, the Gothic Cologne Cathedral (apologies to all Germans - just my esthetic, personal opinion).
Allowing for differences in stretching height, those chairs are the same size in both images, and would be if this shot were taken with a 600mm lens on a huge format camera or if several images were stitched from a smaller format camera, discounting effects of lens distortion.
[attachment=23439:crops_of...ted_0871.jpg]
-
Sorry, it just doesn't work this way. You would need to change the postion of the camera to change the perspective. Changing focal length or sensor/film size doesn't change perspective. (See multiple threads discussing this on this and nearly every other photographic forum)
Yes, you can change the position of the camera, but large format cameras also allow movements that change the relationship of the lens to the film plane in order to allow one to change perspective.
-
But what some people fail to realise, is that the choice of focal length may affect the viewer's impression of the distance to the nearest subject in the composition. The choice of focal length doesn't affect the spatial relationships between the nearest point and the furtherst point in any particular crop, but it does affect the sense of distance to the nearest point (in the composition) to the viewer.
This statement doesn't really make any sense. I'm not sure what you're trying to say when you talk about "impression of distance", but I'm guessing what you mean is that you can get closer to the foreground subject while still including more of the distant ground. But this isn't a function of focal length, it's a function of field of view. That's a significant distinction to make, although it's sometimes lost on people who don't have experience shooting multiple formats. Shooting a scene with 14mm FF DSLR and again with equilvalent FOV on 8x10 LF will give you the same results, focal length has got absolutely nothing to do with perspective or "impression" of perspective. The 14mm shot will probably have barrel distortion, but that has nothing to do with perspective and can be easily corrected. So elf's statement is correct.
-
Yes, you can change the position of the camera, but large format cameras also allow movements that change the relationship of the lens to the film plane in order to allow one to change perspective.
But this has nothing to do with the size relationship of foreground/background. Your previous statement implied that focal length played a role, and it doesn't.
And as far as managing perspective goes, you can do the same thing with a tilt/shift lens or with pano stitching software.
-
Allowing for differences in stretching height, those chairs are the same size in both images,
Ray, the stretching height differences cannot be ignored. I still see a difference of scale in the chairs and tables between your two crop images. Also, the door and windows on the ground floor are at completely different scales in each image as well (what you refer to as stretching height), whereas the scale of the statue/fountain does not appear to change scale as much. It is the perception of these scale differences that I see as being unnatural, but I guess it could be claimed to be subjective.
-
But this has nothing to do with the size relationship of foreground/background. Your previous statement implied that focal length played a role, and it doesn't.
And as far as managing perspective goes, you can do the same thing with a tilt/shift lens or with pano stitching software.
I only indicated that a longer focal length would need to be used for any larger format in order to achieve the same angle of view.
Even if using a tilt/shift lens, you would need a longer focal length in order to end up with the same angle of view in the final stitched image. Tilt/shift lenses can achieve some perspective control, but typically the movements are confined to a single axis (e.g., typically they cannot tilt and swing at the same time).
-
This statement doesn't really make any sense. I'm not sure what you're trying to say when you talk about "impression of distance", but I'm guessing what you mean is that you can get closer to the foreground subject while still including more of the distant ground. But this isn't a function of focal length, it's a function of field of view. That's a significant distinction to make, although it's sometimes lost on people who don't have experience shooting multiple formats. Shooting a scene with 14mm FF DSLR and again with equilvalent FOV on 8x10 LF will give you the same results, focal length has got absolutely nothing to do with perspective or "impression" of perspective. The 14mm shot will probably have barrel distortion, but that has nothing to do with perspective and can be easily corrected. So elf's statement is correct.
Okay, I'll try again to explain what I mean. Let's consider your following statement.
But this isn't a function of focal length, it's a function of field of view...........
Shooting a scene with 14mm FF DSLR and again with equilvalent FOV on 8x10 LF will give you the same results, focal length has got absolutely nothing to do with perspective or "impression" of perspective.
What gives you an equivalent FoV on 8x10 LF? Certainly not a 14mm lens. More like a 100mm lens. No?
Even if you have a variety of different formats of cameras, you often cannot use the same lens on a larger format in order to increase the FoV because the image circle may not be of sufficient diameter. I'd love to be able to use on my 5D one of those Zuko 9mm lenses designed for the Oly 4/3rds format, but I'm pretty sure even if I succeeded in fitting one to my 5D body I'd see very strong vignetting and resolution fall-off towards the edges of the 35mm frame.
Now I agree in an ideal world where we had a perfect lens of fixed focal length, with a huge image circle with 100% MTF at extremely high resolution across the entire image circle, and where such a lens could be used across the entire gamut of different formats of camera, from P&S to 11"x14" large format, then it might be true to say that FoV is not a function of focal length, period, until someone designed another lens of a different focal length which could provide either a wider or a narrower FoV than was previously possible.
Since photographers are making practical decisions every day to change FoV by changing the FL of the lens, why not recognise this practical reality?
However, I agree there is an important principle here. If two images taken from the same position are cropped to the same FoV, whether such cropping takes place in-camera or outside the camera, then the spatial relationships of the same objects within each image will be the same, whatever the focal length used, provided other factors such as focussing are also the same. I would hesitate to claim that an image with fuzzy rocks in the foreground, but sharp cliffs in the background, produces the same sense of perspective in the viewer as a similar scene with tack sharp rocks in the foreground, but fuzzy cliffs in the background, just as I would hesitate to claim that the following two images have the same perspective, even though I know they were both taken with the same lens, at the same time, from the same position.
[attachment=23441:Different_crops.jpg]
Jeff,
On reflection, perhaps I could have expressed this more succinctly and precisely. I'll put it this way. If one changes the focal length of lens but keeps all other image attributes the same, including FOV, focussing, DoF and shooting position, then the sense of perspective should also be the same. This is really self-evident and cannot be argued against. I merely point out that in practice, these other factors often change with changing focal length. The focal length of the lens may also affect the lens design limitations. Is there any wide-angle 100mm or 150mm lens for 8x10 LF that allows for focussing at such close distances as is possible with a 14mm on a DSLR?
-
On reflection, perhaps I could have expressed this more succinctly and precisely. I'll put it this way. If one changes the focal length of lens but keeps all other image attributes the same, including FOV, focussing, DoF and shooting position, then the sense of perspective should also be the same. This is really self-evident and cannot be argued against. I merely point out that in practice, these other factors often change with changing focal length. The focal length of the lens may also affect the lens design limitations. Is there any wide-angle 100mm or 150mm lens for 8x10 LF that allows for focussing at such close distances as is possible with a 14mm on a DSLR?
Actually, the short answer to this is yes, nearly every lens on LF (even 20x24) can focus closer than a 14mm on a DSLR. The focus distance is controlled by the amount of bellows draw. I saw a for sale ad last year for an 8 foot long bellows setup for 8x10:)
It's probably fair to say that most people don't stitch or crop to match another format and for that matter the vast majority of people don't stitch, so they don't see that it (edit: perspective) doesn't change except when the camera position is moved. It's hard to explain perspective when 95% of the time perspective is described in terms of getting the same FOV from different formats.
p.s. I like the uncorrected version much better. The leaning/distorted buildings give the squatty ladies something to be distressed about
-
Since photographers are making practical decisions every day to change FoV by changing the FL of the lens, why not recognise this practical reality?
Ray,
What you say is effectively true in the context of a single format such as 35mm (unless you bring stitching into the picture).
My preference for talking about FOV rather than focal length might seem like splitting hairs from a practical standpoint. But the problem is that when you talk about focal lengths, some people get confused and start to ascribe characteristics of FOV and perspective to focal lengths. This can lead to people saying stuff like "I prefer full-frame over DX, because my 85mm lens on FF gives a more pleasing perspective for portraits than 60mm on DX"; or claiming that you can't get the same perspective as a wide-angle shot by stitching multiple shots with a longer lens.
-
p.s. I like the uncorrected version much better. The leaning/distorted buildings give the squatty ladies something to be distressed about
I like the idea. I might make a large print of the uncorrected image. The more allusions the better. Photos should not only be about illusions, but also allusions.
There are now at least two things for these women to moan about. Yet another church which is going to increase their tax burden, and dangerously leaning buildings which are about to collapse any moment.
(I understand in Germany people have to pay additional taxes to support their churches, unless they take the trouble to inform the taxation office that they are not Christian. I might have got this wrong, but that's what I've heard.)
-
Ray,
What you say is effectively true in the context of a single format such as 35mm (unless you bring stitching into the picture).
My preference for talking about FOV rather than focal length might seem like splitting hairs from a practical standpoint. But the problem is that when you talk about focal lengths, some people get confused and start to ascribe characteristics of FOV and perspective to focal lengths. This can lead to people saying stuff like "I prefer full-frame over DX, because my 85mm lens on FF gives a more pleasing perspective for portraits than 60mm on DX"; or claiming that you can't get the same perspective as a wide-angle shot by stitching multiple shots with a longer lens.
Yes, I understand that's why some of you are so adamant that FL has nothing to do with impressions of perspective when considered as an isolated factor. However, because most cameras produced nowadays are miniature P&S cameras with typical focal lengths ranging from 4mm to 15mm, with a 50mm lens acting like a long telephoto, there should be no confusion that focal length of itself can have any direct bearing on the sense of perspective in an image.
My point really is just that a camera of a specific format is always required to take a photo, in addition to a lens, and therefore such format of camera should always come into the equation because it may determine the selection of focal length of lens for a desired FoV. The FoV in the final image, which may affect the impression of perspective from the standpoint of the person viewing the image, cannot necessarily be claimed to have nothing to do with the focal length of lens, because the choice of FL of lens may have determined the FoV. They are related in practice.
Likewise, those photographers who insist on using one fixed lens for all their shots will find that it tends to change the perspective in at least some of their images because they have to use their legs more to get closer to the subject, instead of being lazy and using a zoom from a greater distance. There are indirect perspective effects flowing from the choice of lens focal length, just as there may be indirect perspective effects created by the greater FOV surrounding a particular subject as a result of the use of a wider angle of lens.