Luminous Landscape Forum
Equipment & Techniques => Landscape & Nature Photography => Topic started by: julian kalmar on March 21, 2007, 11:55:59 am
-
(http://gallery.photo.net/photo/5657015-lg.jpg)images original size: 18762x9490
For technic freaks: I have been asked very often, wy I prefer stching instead of making 4x5 or 8x10 larg format. During my studies I did a lot with 4x5 inch and even some 8x10 inch work. The answer is very simple: Stiching 3 or 4 images is very simple and the resolution is much better then 6x7 medium format ( with my 5d) stiching 10 images gives you a bit better resolution then 4x5 inch and still this is not a big game compared with taking a 4x5 inch gear with you. stiching 20 or more images gives you a much better resolution then 8x10 inch but stiching becomes complicate but the result is breathtaking Remember: I only talk about resolution, nothing else (I do not whant to start that stupid conversation if film ordigital is better Many amateurs think that larg format has 10 times more resolution then 35mm. IT HAS NOT!! Why? Well, the answer is simple: the lens has to fill a bigger image circle, and the bigger the circle , the less is the resolution of the lens. If you do not believe, try to make a 100 inch print from a 8x10 inch slide. If you make a 100 inch print from a stiched image with 45 images you get a native resolution of 200 dpi and the image is rasor sharp, even if you look from 30 inch distance
For stiching I used PTGui and the seperate layers I put together in PS
You are alsow welcome to visite my Gallery: http://photoart.lima-city.de/ (http://photoart.lima-city.de/)
(http://photoart.lima-city.de/album1280x1024/Pano/slides/IMG_4461-5508%20555%2065-68%2078-91.jpg)
I forgot to post a 100% crop sorry http://gallery.photo.net/photo/5657039-lg.jpg (http://gallery.photo.net/photo/5657039-lg.jpg)
-
very impressive, how long would it take to create that with the software?
-
very impressive, how long would it take to create that with the software?
[a href=\"index.php?act=findpost&pid=108042\"][{POST_SNAPBACK}][/a]
I stiched with PTGui but I had lots of mistakes ( especially in the sky)
So I made seperate layers ( and this last much longer then creating the complete image ) and put the pacle together in PS. I think my computer worked about 3-4 hours (2,8GHz HT 1GB Ram)
Totally this image consists more than 160 shots ( I had to take each image 3-4 times with different exposures and focusing points) From these images I made 45 "perfect" images for stiching
Of course in half an hour the sky was chanching a lot which resulted in problems for stiching. For putting the pacle together I needed 2 month
-
Very impressive image! It would be difficult to obtain the images you needed, and also have the lighting reasonably the same in all of them, and before the clouds moved.
What kind of tripod head do you use? Something that has clickstops at each horizontal and vertical interval is probably useful when there are this many images in a grid that must be taken.
Andy
-
The perspective is soo unnatural...I don't find it attractive at all.
-
It is an impressive stitch. If printed Clyde Butcher big it might be neat but not so much at this size.
-
Great pix, Julian!
With years of hands-on with up to 8x10 film capture and now nearly all digital, I, too, share your observation that digital can equal and in many cases, surpass the resolution of film.
I have on my wall a 12-shot pano of Factory Butte in Utah that would be a royal PIA had I shot it with my 10 inch Cirkut camera. Not only the quality is better, but just processing and printing that 10 inch by 6 feet film was always an adventure and made the darkroom a mess!
With the new image merge in CS3, it's even easier and better than hand stitching!
We're still in the horse-and-buggy era of digital and this will only get better!
I'm printing about 100 prints, mainly 20x30 to 24x26 and longer from the panos right now that rival anything I once got with film. Even the high-res scans from our 645 aren't nearly as sharp as the stuff from our D200 and D2x captures and it takes the added step of scanning.
When you get down to the nuts and bolts, the main reason for large-format in the first place was that was the only way to make large prints at one time. Granted, the lenses today for medium and large format have substantially better than they were in the golden age of both just a few years ago, but Nikon, Canon and the rest did there homework and pushed the envelope. They had to.
If you could get the same image quality from a lens on MF or larger that we get on the current crops of 35 and DX lenses, then there would be no point in shooting the smaller formats. But in reality, there isn't much market to make those lenses.
In the overall scheme of things, though, my Zeiss lenses on the 'blad are awfully nice, and it's hard to get that image quality on anything smaller, but with stitching, it's getting harder and harder to want to shoot film.
Look at Steve Johnson's work with both one-shot and scanning backs. He's on the bleeding edge and is pushing digital limits way beyond what most of us will ever do. But with with what most people have in hand and on their desktops, a little skill and craftsmanship can go a long way in getting into Steve's league.
Go out and shoot some more!
-
When you get down to the nuts and bolts, the main reason for large-format in the first place was that was the only way to make large prints at one time.
[a href=\"index.php?act=findpost&pid=108123\"][{POST_SNAPBACK}][/a]
Gosh, you mean I've been fussing with movements for nothing?
-
Great Image. I have been there and feel you captured the place.
Charlie
-
What kind of tripod head do you use? Something that has clickstops at each horizontal and vertical interval is probably useful when there are this many images in a grid that must be taken.
Andy
[a href=\"index.php?act=findpost&pid=108090\"][{POST_SNAPBACK}][/a]
Thank´s for all the interests! I´m using a self build Pano head (those you can buy are either very expensive and heavy, or cheap but not able to hold my camera)
I don´t use clickstops because I´m faster when I look in my camera ( It´s very simple: I´m just looking that the seoerate images are crossing each other about one third, so I´m always on the save side)
sorry for my bead English, I hope you are able to understand what I mean.
-
The perspective is soo unnatural...I don't find it attractive at all.
[a href=\"index.php?act=findpost&pid=108103\"][{POST_SNAPBACK}][/a]
Doug,
You are the re-incarnation of a diplomat, aren't you?
Cheers,
Bernard
-
Doug,
You are the re-incarnation of a diplomat, aren't you?
Cheers,
Bernard
[a href=\"index.php?act=findpost&pid=108732\"][{POST_SNAPBACK}][/a]
You know, from the picture it looks like two diplomats have been squeezed in there!
It was a very sour comment, wasn't it?
John
-
Very impressive - esp. the fact that you included the people for scale is even more impressive!
-
I agree with the stated benefits of stitching and do it myself regularly... However, WRT LF lens resolutions, you are repeating old lens data, not data from the current state of LF lenses. I have 4x5 and 8x10 lenses that I have tested at 60 line PAIRS per mm -- and this is on par with the best Canon glass I own. I suspect some MF and 35mm lenses will do better, but even the small pixels on the D2x or 30D sensor are tapped out by about 70 line pairs when the AA filter is factored in. The other advantage you miss is the movements available on the view camera are useful. Bottom line is there a point of diminishing returns for stitched images; certainly a 4 or 6 frame capture can be captured quickly and assembled quite easily for a superb result, but over a dozen and things start to get complicated really quickly. So IMO film is far from left out. All you need to do is stick your nose in one of Burtinsky's 60" prints next time his exhibit is in your neighborhood -- talk about detail! It'll make you want to run out and buy an 8x10 camera...
Cheers,
-
There is a curious effect in this image which seems a bit odd. The distant horizon, top of the canyon, doesn't look distant. Perhaps this is due to unfamiliarity with such sharp images. One doesn't expect distant objects to be razor sharp.
-
There is a curious effect in this image which seems a bit odd. The distant horizon, top of the canyon, doesn't look distant. Perhaps this is due to unfamiliarity with such sharp images. One doesn't expect distant objects to be razor sharp.
[a href=\"index.php?act=findpost&pid=108796\"][{POST_SNAPBACK}][/a]
Good call Ray. I didn't know what it was that didn't look quite right, but there is a curious lack of depth for that type image. Of course it might just have been a really clear day.
(FWIW, it also appears by looking at the crop that the file was over-sharpened on conversion or at least in post.)
Cheers,
-
This is a point I have tried to figure out for myself - but I am not sure. Someone much smarter then me must have thought this through.
I think it is well accepted that using a tele lens tends to give a picture that has a compressed 'Depth of Field Look'. And using a very wide angle lens can increase the preceived 'Depth of Field Look'.
Therefore, if you take a single picture of Bryce Canyon with a 24mm lens it will have a large 'Depth of Field Look'. But when you take many pictures with say a 180mm lens and stich them together to get the same angle of view, the same total picture - won't it have the compressed 'Depth of Field Look'?
And isn't that a problem with stiching in general?
Your sane comments on my rational would be appreciated.
And, oh yes, I really like the picture at the start of the thread plus many others on his web site.
Thanks, Dave
-
Therefore, if you take a single picture of Bryce Canyon with a 24mm lens it will have a large 'Depth of Field Look'. But when you take many pictures with say a 180mm lens and stich them together to get the same angle of view, the same total picture - won't it have the compressed 'Depth of Field Look'?
[a href=\"index.php?act=findpost&pid=108893\"][{POST_SNAPBACK}][/a]
Dave, I think if you make the stitched 180mm shot have the exact same Angle Of View and take it from the exact same position as the single 24mm shot, then the overall persepctive will be identical in both and in fact the only differences will be image detail... It gets sticky with Depth Of Field, because we are using a longer lens, but also creating a larger net file, so the two factors kind of cancel each other out for any given aperture. (IOW if we used f8 on both lenses, the net DOF should be similar in both.) Hence, neither image should have more or less apparant "depth" than the other due to perspoective or DOF.
HOWEVERBUT! I think the added detail does in fact alter our sense of depth as we can see minute detail structure in every part of the stitched image, but cannot say the same thing for the single 24mm capture. So perhaps that is what makes the stitch appear to have less depth?
Cheers,
-
It gets sticky with Depth Of Field, because we are using a longer lens, but also creating a larger net file, so the two factors kind of cancel each other out for any given aperture. Cheers,
[a href=\"index.php?act=findpost&pid=108918\"][{POST_SNAPBACK}][/a]
Yes, I think that's right. Take 4 overlapping shots with a 35mm camera (2 rows of 2 images) and the result is very similar to a single shot with a FF 6x4.5cm format using the same focal length of lens. Take a mosaic of around 64 images with a 35mm camera, using a 350mm lens, and the result would be very similar to a single shot with an 8x10" field camera, using a 350mm lens.... except the stitched image would be much sharper.
-
The distant horizon, top of the canyon, doesn't look distant. One doesn't expect distant objects to be razor sharp.
[a href=\"index.php?act=findpost&pid=108796\"][{POST_SNAPBACK}][/a]
Dear Ray,
The answer is very simple. The top of the Canyon is not very distant. Look on the very left side of the image behind the Canyon. That is distant
-
Dave, I think if you make the stitched 180mm shot have the exact same Angle Of View and take it from the exact same position as the single 24mm shot, then the overall persepctive will be identical in both and in fact the only differences will be image detail... It gets sticky with Depth Of Field, because we are using a longer lens, but also creating a larger net file, so the two factors kind of cancel each other out for any given aperture. (IOW if we used f8 on both lenses, the net DOF should be similar in both.) Hence, neither image should have more or less apparant "depth" than the other due to perspoective or DOF.
HOWEVERBUT! I think the added detail does in fact alter our sense of depth as we can see minute detail structure in every part of the stitched image, but cannot say the same thing for the single 24mm capture. So perhaps that is what makes the stitch appear to have less depth?
Cheers,
[a href=\"index.php?act=findpost&pid=108918\"][{POST_SNAPBACK}][/a]
Thank you Jack. And, yes I agree one has to be aware of using a higher mm lens with the 'Depth of Field' (focus) being less at the same f stop. Usually I do increase the f stop used because of that.
Thanks again.
-
Julian,
You talk about two months of computer time assembling this image. How much time elapsed in the field making the 160 captures?
I routinely use stitching of 2-4 images to solve problems with commercial shoots, but for anything like your example I would go to Large Format in a minute. In retrospect, you really feel this is a superior method to shooting Large Format? I can see it for someone who does not own large format, but for me LF for this kind of image would be a no brainer for both capture in the field and image processing. Hours in the field become a few minutes. Months in front of the computer becomes a few hours.
-
Julian,
You talk about two months of computer time assembling this image. How much time elapsed in the field making the 160 captures?
I routinely use stitching of 2-4 images to solve problems with commercial shoots, but for anything like your example I would go to Large Format in a minute. In retrospect, you really feel this is a superior method to shooting Large Format? I can see it for someone who does not own large format, but for me LF for this kind of image would be a no brainer for both capture in the field and image processing. Hours in the field become a few minutes. Months in front of the computer becomes a few hours.
[a href=\"index.php?act=findpost&pid=108976\"][{POST_SNAPBACK}][/a]
1: sorry! I do not have a larg format lens with 125 degree field of view. Do you?
2: Just believe me: I can get more detail with stiching 45 images and I needed the details for a 7 feet print, that´s why decide to stich these image
-
except the stitched image would be much sharper.
[a href=\"index.php?act=findpost&pid=108951\"][{POST_SNAPBACK}][/a]
Ray: I'll give you everything except this last comment. As I said above, I have 360mm LF lenses that can lay down over 60 line pairs at the film.
Cheers,
-
Julian, taking also images for stitching with a 5D I correct my source pics with PTLens (distortion, vignetting) and stitch them with Autopano, works great.
-
Julian,
So rent a view camera and do a 2x with 8x10 or 4x stitch with 4x5? The time involved is still a small fraction of the time you have expended here.
I do this professionally and do stitches from both DSLR's and 4x5 all the time for clients. To me, while your effort is an interesting exercise, it is not a viable routine technique for a variety of reasons.
-
Dear Ray,
The answer is very simple. The top of the Canyon is not very distant. Look on the very left side of the image behind the Canyon. That is distant
[a href=\"index.php?act=findpost&pid=108954\"][{POST_SNAPBACK}][/a]
Dear Julian,
The top of the Canyon is as distant as the trees are small, relative to the closer figure lower down. The elevation of the camera seems to be about midway between the top of the Canyon and the bottom. For some reason, the distance between the small figure, lower right, and the trees on top of the ridge (with regard to depth rather than elevation), seems a bit odd.
Of course, I've never visited this location. That might be exactly how it looks. Perhaps also, on a small jpeg vital clues as to depth are missing. Perhaps the stitching program has distorted the size of certain elements. I just mention it as an impression I get. Impressive image though .
-
Ray: I'll give you everything except this last comment. As I said above, I have 360mm LF lenses that can lay down over 60 line pairs at the film.
Cheers,
[a href=\"index.php?act=findpost&pid=108981\"][{POST_SNAPBACK}][/a]
Jack,
That is indeed impressive. Are we talking 4x5 or 8x10? And what film and at what MTF?
I believe T-Max 100 (is it still available?) can (could) record 50 lp/mm at 100% MTF and 100 lp/mm at 60% MTF. I guess it could record 60 lp/mm at about 85% MTF. A good 360mm lens for 35mm format should be able to deliver 60 lp/mm to the film at 50% MTF.
50% of 85% is 42% MTF. I suspect those 60 lp/mm from your LF 360mm lens are barely discernible; 10% MTF or less. Am I right?
-
Jack,
That is indeed impressive. Are we talking 4x5 or 8x10?
~~~
I suspect those 60 lp/mm from your LF 360mm lens are barely discernible; 10% MTF or less. Am I right?
[a href=\"index.php?act=findpost&pid=109062\"][{POST_SNAPBACK}][/a]
Ray:
All of the LF 360's I've owned will cover 8x10, so we're talking 8x10 I did not measure contrast, but the line-pair divisions were clearly visible (no guessing), so I'd estimate more like 25% or 30%. And admittedly, film grain adds to one's ablity to distinguish detail. Regardless of whatever the contrast was, there was a clearly visible distinction on the line pairs -- just like what I'd look for to determine "real" detail in an image. This was in the central area of the IC, and using the same arbitrary yet "clearly visible distinction," the lenses made over 50 at the extreme corner of the 8x10 frame WITHOUT re-focusing to compensate for field curvature.
FWIW, I also tested several different 4x5 lenses when I had my Betterlight scanning back. Here, many of the contemporary pieces of glass equalled the resolution limits of the sensor, about 54 line pairs. Even some old lenses from the early 1930's through the 1950's got very close to that, though many tapped out at around 30 line-pairs. Surprised me actually to discover that as a group, the contemporary large format lenses were that good...
Cheers,
-
FWIW, I also tested several different 4x5 lenses when I had my Betterlight scanning back. Here, many of the contemporary pieces of glass equalled the resolution limits of the sensor, about 54 line pairs. Even some old lenses from the early 1930's through the 1950's got very close to that, though many tapped out at around 30 line-pairs. Surprised me actually to discover that as a group, the contemporary large format lenses were that good...
[a href=\"index.php?act=findpost&pid=109071\"][{POST_SNAPBACK}][/a]
Yes, it is surprising. I still think, though, that the MTF of those 60 lp/mm is critical for real world images. Even cheap 35mm zooms can deliver 60 (and more) high contrast line pairs which are still easily distinguishable even though they might have lost 80 or 90% of their original contrast by the time they are recorded.
What happens if those line pairs are not high contrast, say medium to low contrast? A camera like the Canon 20D, although its sensor is capable of recording 60 lp/mm, simply wouldn't be able to record them using a cheap zoom. But it probably would when using a good prime, provided the contrast of the target wasn't t too low.
I would find it difficult to believe, if I were to use my modestly okay Canon 100-400 IS zoom at 400mm and f16 on my 5D, and take a mosaic of around 64 images which, when stitched, would have the same FoV as a single shot with an LF 400mm prime, that the single 8x10" shot at f16 would even be nearly as sharp, assuming you didn't use shift movements.
I could be wrong, though. Would someone care to do the experiment .
-
Yes, it is surprising. I still think, though, that the MTF of those 60 lp/mm is critical for real world images. Even cheap 35mm zooms can deliver 60 (and more) high contrast line pairs which are still easily distinguishable even though they might have lost 80 or 90% of their original contrast by the time they are recorded.
What happens if those line pairs are not high contrast, say medium to low contrast? A camera like the Canon 20D, although its sensor is capable of recording 60 lp/mm, simply wouldn't be able to record them using a cheap zoom. But it probably would when using a good prime, provided the contrast of the target wasn't t too low.
I would find it difficult to believe, if I were to use my modestly okay Canon 100-400 IS zoom at 400mm and f16 on my 5D, and take a mosaic of around 64 images which, when stitched, would have the same FoV as a single shot with an LF 400mm prime, that the single 8x10" shot at f16 would even be nearly as sharp, assuming you didn't use shift movements.
I could be wrong, though. Would someone care to do the experiment .
[a href=\"index.php?act=findpost&pid=109079\"][{POST_SNAPBACK}][/a]
Don´t missanderstand me I didn´t whant to start those redicolous debate if film or digital is better. I´m still using my sinar 8x10 and have wonderful lense ( 121mm super Angulon, sironar 240mm Apo sinaron 360mm) Of course I know that you can´t stich everything. But to be honest : Pano shooting is not a domain for 8x10 inch ( you always have problems with contrast and I know no Pano had which can hold my camera and I know no computer which is able to stich a scanned 8x10 slide. If you don´t whant to loose detail you have to scan it with high resolution and 16 bit and you get a 2GB file for each image)
What I like on film (8x10) is that you have to compose your picture wheras with digital you just click it. I mean no one can "compose" a 45 image stich made of more than 160 images you never no if it works in the end.
Now the resolution debate:
The best lenses for 35mm are able to repoduce 135 line pairs per mm in resolution charts. This would result in a 5940 x 9180 pixle image. 54 MP!!!
Everyone knows that this is nonsense!! You have to be glad if you get 12 MP in real world photography ( normaly 8-10 MP) So you only get one fourth of the resolution the chart shows. Why? The answer is simple: you normaly don´t make photos from linepairs which have a contrast of 1:1000 and you loose this resolution very quickly with decreasing of contrast because film grain whips it
-
Now the resolution debate:
The best lenses for 35mm are able to repoduce 135 line pairs per mm in resolution charts. This would result in a 5940 x 9180 pixle image. 54 MP!!!
Everyone knows that this is nonsense!! You have to be glad if you get 12 MP in real world photography ( normaly 8-10 MP) So you only get one fourth of the resolution the chart shows. Why? The answer is simple: you normaly don´t make photos from linepairs which have a contrast of 1:1000 and you loose this resolution very quickly with decreasing of contrast because film grain whips it
[a href=\"index.php?act=findpost&pid=109088\"][{POST_SNAPBACK}][/a]
That's true, but we've been talking about 60 lp/mm. These figures are not nonsense.
Also, there are parts of many images that do have high contrast material, such as leaves against a sky, trees on a horizon, car number plates, signs and billboards in cities etc.
The general principle is, a line pair specification is meaningless unless it's associated with an MTF percentage.