Luminous Landscape Forum
The Art of Photography => The Coffee Corner => Topic started by: PeterAit on May 26, 2015, 03:05:48 pm
-
http://www.telegraph.co.uk/travel/snowandski/france/11628359/365-gigapixel-Mont-Blanc-panorama-becomes-the-worlds-largest-ever-photograph.html
-
I don't really understand the point. Is this one of those "let's make the world's largest hotdog!" things?
-
Completely pointless, but still cool. This was shot with a Clauss Pixpert head it would seem.
Good stuff, I happen to own one of those. ;)
Cheers,
Bernard
-
Quite a head, considering it was attached to a 400mm f/2.8 lens and camera.
Still not a record, merely a record for terrestrially based cameras. NASA holds the record at some 600 GP for a Moon map.
-
Completely pointless, but still cool. This was shot with a Clauss Pixpert head it would seem.
Good stuff, I happen to own one of those. ;)
Cheers,
Bernard
I just took a look on their page. Proud prices :o
-
How much RAM do you need to pp THAT?
-
Interesting record - boring picture...
I know the Mont Blanc is usually in the clouds so good timing
..on the other hand when it is not in the clouds - boring weather...
-
Quite a head, considering it was attached to a 400mm f/2.8 lens and camera.
Yes, I use it with a Nikon 400mm f2.8 E FL too and it does perform pretty well.
Cheers,
Bernard
-
Completely pointless
Sure?
-
Sure?
Yes. 300 megapixel makes a lot of sense compared to 50~80, but 300 gigapixels is pointless.
Cheers,
Bernard
-
Yes. 300 megapixel makes a lot of sense compared to 50~80, but 300 gigapixels is pointless.
Cheers,
Bernard
ever tried to stich 70.000 photos on your computer?
-
ever tried to stich 70.000 photos on your computer?
No, the most I have done is around 200 I guess.
As mentioned, I don't see any value in going beyond 1 giga pixel. I have better things to do in life than doing stuff I believe doesn't make sense. ;)
Cheers,
Bernard
-
No, the most I have done is around 200 I guess.
As mentioned, I don't see any value in going beyond 1 giga pixel. I have better things to do in life than doing stuff I believe doesn't make sense. ;)
Cheers,
Bernard
exactly: you believe it doesn't make sense.
from a technical point of view it does a lot of sense.
-
Well for me it one of those Guinnessbook records....
What it brings is publicity. the making of ... etc etc...
The image itself is not even stitched well- difficult with snow?
-
The image itself is not even stitched well- difficult with snow?
Poor stitching or the limitations of the time it takes to take 70,000 images?
-
Well for me it one of those Guinnessbook records....
where you see just a stunt I sent a huge technical challenge.
just think how difficult can be to produce a motor head that keeps its precision at low temperatures, or the algorithm needed to stitch 70.000 full res shot.
-
where you see just a stunt I sent a huge technical challenge.
just think how difficult can be to produce a motor head that keeps its precision at low temperatures, or the algorithm needed to stitch 70.000 full res shot.
Do not know about the motorhead, but the stitching challence is not that great since even i could do it.
It is just more, more and more...
-
Do not know about the motorhead, but the stitching challence is not that great since even i could do it.
It is just more, more and more...
really you can stitch 70.000 full resolution photos?
on what computer?
-
really you can stitch 70.000 full resolution photos?
on what computer?
On any- if you have a big problem divide it into parts... old computers may just take a bit longer and need smaller parts..
The accuracy they have reached on stitching suggests that the result is in some way closer to 18 Gigapixel than 365Gigapixel
Update
the quality of the outcome could have been better if they had used a 36Mp Nikon or the 50MP Canon with a 300mm lens- this means less stitches and less time ( so less problems)
or using the same 400mm lens and turn the record into about 500-700 Gigapixel
-
kers,
The image itself is not even stitched well- difficult with snow?
Given the level of detail in the image in question, no. The result is due to poor control point placement, albeit cablecar motion cannot have helped if it existed. Unfortunately the enormous number of images required for such images means that accurate control point placement by manual means is impracticable.
Diego,
...just think how difficult can be...the algorithm needed to stitch 70.000 full res shot.
I suspect the algorithm used to stitch two images is identical to that used to stitch 70,000. CPU speed, temporary and permanent memory etc. will be the principal limiting factors.
-
...
Unfortunately the enormous number of images required
...
I wonder why they didn't just divide the problem in smaller parts.
-
Diego,
I wonder why they didn't just divide the problem in smaller parts.
Because manual control point placement would still be required for 70,000 images.
-
Diego,
Because manual control point placement would still be required for 70,000 images.
I'm not the one suggesting that "dividing into smaller parts" solves all the problems in this kind of processing.
-
Diego,
I'm not the one suggesting that "dividing into smaller parts" solves all the problems in this kind of processing.
Who do you consider is suggesting that "dividing into smaller parts" solves all the problems in this kind of processing?
-
Diego,
Who do you consider is suggesting that "dividing into smaller parts" solves all the problems in this kind of processing?
http://forum.luminous-landscape.com/index.php?topic=100667.msg826477#msg826477
-
Diego,
In the post you linked to, kers does not claim that "dividing into smaller parts" solves all the problems in this kind of processing. kers addresses your question, which relates to one problem...in this kind of processing.
-
Diego,
In the post you linked to, kers does not claim that "dividing into smaller parts" solves all the problems in this kind of processing. kers addresses your question, which relates to one problem...in this kind of processing.
Sure?
You wrote: "the enormous number of images required for such images means that accurate control point placement by manual means is impracticable".
Not impossible: impracticable.
Stitching 70.000 images on a 386 with 4Mb of RAM is not impossible: it's impracticable.
So saying that "any computer can stitch 70.000" is as correct as saying "control point placement by manual means on 70.000 images can be done manually".
Unfortunately, sometimes size does matter.
-
hai Diego and Rob,
Thank you both for the remarks,
I just saw on dpreview :
"A team of five photographers has created what they describe as the world's highest definition panoramic photograph by stitching 70,000 digital images together to create a 365 gigapixel photograph. Recorded using a Canon EOS 70D with the EF 400mm f/2.8L IS II USM lens on a 2x converter, the picture took 35 hours of shooting over the course of 15 days just to capture. A further two months were required in post-production to stitch the images together, creating a 46 terabyte finished product."
So in effect a 1200mm lens. I see now that badly stitched part is only a small problem - the static mountain looks good and well stitched.
Now i think it is a problem to even put the 46 terabyte in your computer, so it simply had to be divided in smaller parts resulting in a zoomable website - the real end product.
Like a google map- tiles are loaded that you need. Coarse tiles and smaller tiles..depending on the zoom level. There is no complete 46 tb photo anywhere, but the building blocks are there to make it.
-
I bet LR6 could handle the stitch no problem.
-
hai Diego and Rob,
....
so it simply had to be divided in smaller parts resulting in a zoomable website - the real end product
....
How simple it is?
-
stamper,
I bet LR6 could handle the stitch no problem.
Certainly PTGui can stitch as many as 65,536 images on a Mac, according to section 3.18 of its FAQ (http://www.ptgui.com/support.html).
-
How simple it is?
i do not know exactly,
The technique is used on all digital maps... i am not an expert, but the technique is there to be used.
Some parts of the image are truly badly stitched - the MontBlanc part is mostly very good..of course.. except for the cabin
The photo is taken from the Italian side- The cabin you see is really nice to take if you are there.
It is purely for the sightseeing- takes 20 minutes and stops now and then to give you a quiet view.
It connects italy and France- moves horizontally and is made by the Suisse if i remember correctly.
Halfway it is only supported once. I have a big picture of the view from it hanging on my wall that i took with a 4x5 inch and i can see even more detail on this panorama...
So must say they photographed the mountain as big as it really is.
Since i do not see any moved or vibrated images, the weather must have been very good and not to much wind.
-
i do not know exactly,
The technique is used on all digital maps... i am not an expert, but the technique is there to be used.
Are you sure that the same technique that works with 200 images still works with 70.000 images?
A "some terabyte" database can run on a single machine: google databases need multiple datacenters.
-
Are you sure that the same technique that works with 200 images still works with 70.000 images?
A "some terabyte" database can run on a single machine: google databases need multiple datacenters.
not sure no, but maybe you have an answer here:
https://openseadragon.github.io
it is the javascript engine they used.
cheers PK
-
not sure no, but maybe you have an answer here:
https://openseadragon.github.io
it is the javascript engine they used.
cheers PK
This is the engine to display the "already stitched" tiled image.
The problem is stitching the 70.000 images together.
-
It seems like this photograph is interesting not because of the image but because of the technique used to produce it.
They could have taken 70,000 pictures of a parking lot and still have gotten this "record".
-
It seems like this photograph is interesting not because of the image but because of the technique used to produce it.
They could have taken 70,000 pictures of a parking lot and still have gotten this "record".
the value of this kind of thing lays much more on the failures than the successes: find out what doesn't work and why is more useful.
-
About bad stitching: you good people forget that the "photograph" was taken in a course of a week or so. Impossible to expose all 70000 frames at the same instant. That means there will be places where the light (time of the day) is different in connecting frames. They might be exposed days apart.
Solution for the next, better attempt: a giant fill-in flash to even out the light...
-
It also said the took pictures from different positions which would also explain what look like parallax errors on the cablecar wires.
-
I think we all would care if the photo was a great photograph and something worth capturing. Instead another poor attempt of stitching. Stitching for the sake of stitching does not make a photo, but a massive amount of space. Would be nice if it was taken with great lighting, composition and something at least interesting. T
-
I think we all would care if the photo was a great photograph and something worth capturing. Instead another poor attempt of stitching. Stitching for the sake of stitching does not make a photo, but a massive amount of space. Would be nice if it was taken with great lighting, composition and something at least interesting. T
If there are 70000 frames to expose, you would have to get the Great Lightman to stop earth from rotating and clouds moving to keep the scene still for 5 days…
This endeavor is really not about stitching quality or even about an interesting picture, but to make a new world record.
-
This endeavor is really not about stitching quality or even about an interesting picture, but to make a new world record.
And test the limits of present technology.
-
I prefer the >500 GP Moon image set (NASA) made up from a zillion orbit photos. :D