Follow up on Ben's "Calculate This" thread

[Cem_Usakligil]

Hi folks,

Perhaps some of you have already read the "Calculate this" thread started by Ben. I have actually gone out to shoot a test picture as I have mentioned here. For completeness, here is what I wrote:

Hi Ben,

Glad to read that you are pursuing this further. Inspired by you, I have conducted a similar experiment myself yesterday. I chose a subject similar to yours, a wall with a couple of windows and an open door, which I have photographed at an angle (the distance of the camera to the middle part of the wall was roughly 3 meters). I have shot a single frame with the 24mm @f5.6, and I have shot roughly the same composition using the 100mm @f11. I ended up with 44 frames (due to the 50% overlap) which have then been stitched in a quick and dirty way (i.e. with some minor parallax/blending/stitching errors which need correcting in a final version). It was a very interesting experiment. The resulting image is some 22000x13500 pixels, which would be equivalent to a hypothetical digital sensor size of 141mm x 86mm (297 megapixels) using the same pixel pitch. The dof of the stitched image is lower than the 24mm image, as was expected. Looking at both pictures side by side on my 30" monitor, I can't say that I see an immediate difference in tonality (which was the main goal of the exercise). This should be because of the fact that both images are downsized to display them on the monitor. But I did not scrutinize them long enough yet. The proof of the pudding would be in the printing (at A3+ size or larger). However, since this was only a test picture without value I am not sure I shall bother printing it. Subjectively, the stitched image feels more robust and of course one can zoom into it almost indefinitely; the details are mind boggling. Never mind, I will let you know how and if this continues. If you want me to print your own pictures and send them back to you, I am gladly willing to do so; as I wrote earlier.

I thought it would be a good idea to show you the results. I did not want to pollute Ben's thread, so I am starting another one here. Mind you, the photo is not a work of art or anything like that, it has been shot for test purposes.

First, here is the single frame picture taken using the TSE 24L f3.5 Mk II on my Canon 5D Mk II. This was taken when sun was shining and it has cast some contrasty shadows as you can see. This is a tone mapped picture from 3 bracketed exposures in SNS-HDR.

Next, the stitched picture using the EF 100 macro f2.8. There were 44 frames in total. Each frame was in turn tone mapped from 3 bracketed exposures in SNS-HDR. I have now properly stitched this in PT Assembler and used equirectangular projection. During the shoot, the sun has disappeared behind clouds and I even had to stop shooting due to a sudden rain-shower. There are no direct shadows and the contrast is lower.

You can see that both pictures look quite similar. The stitched version looks even a bit duller compared to the single frame version. They are both resized for showing here and most of the differences have disappeared. Therefore, one has to see the details at 100% pixel level to know the differences.

A 100% crop (1000x800 px) of the single frame 24mm picture.

And next, a 100% crop (1000x800 px) of the 100mm stitched picture, zooming in on the same area as above.

The level of detail in the stitched picture is incredible. I reckon I could print this easily 10ft (3m) wide if I wanted to.

Ask away if you want to know any details.

 

[Asher Kelman]

Next, the stitched picture using the EF 100 macro f2.8. There were 44 frames in total. Each frame was in turn tone mapped from 3 bracketed exposures in SNS-HDR. I have now properly stitched this in PT Assembler and used equirectangular projection. During the shoot, the sun has disappeared behind clouds and I even had to stop shooting due to a sudden rain-shower. There are no direct shadows and the contrast is lower.

Cem,

I'm not clear about the 44 frames. Are you saying that there were actually 132 frames counting the bracketing?

And next, a 100% crop (1000x800 px) of the 100mm stitched picture, zooming in on the same area as above.

The level of detail in the stitched picture is incredible. I reckon I could print this easily 10ft (3m) wide if I wanted to.

This is a remarkable demonstration of not only the benefit of stitching and SNS-HDR, but also of your dedication. If the sum had kept shining, the 3D effect would be amazing. To make use of this level of work, consider output to Cibachrome paper as is has 100 lp/mm resolution and great color.

Asher

 

[Cem_Usakligil]

Hi Asher,

I'm not clear about the 44 frames. Are you saying that there were actually 132 frames counting the bracketing?

Yes indeed, 44 x 3 = 132 frames in total.

 

[Ben Rubinstein]

What were you trying to achieve Cem? What was the end goal?

 

[Cem_Usakligil]

Hi Ben,

What were you trying to achieve Cem? What was the end goal?

As I wrote in your original thread, you have inspired me to shoot in this way. My goals were:
a) Is it doable or not? It was mainly an exercise to fine-tune my stitching skills and to identify any potential pitfalls in case I wanted to use this method in the future. Such pitfalls as in the area of focusing and non parallax rotation, etc.
b) I wanted to see if there would indeed be a tonality difference between the single frame and the large one. And if so, would that be visible unless printed very large or not.

The answer to question a is a yes, it is doable.

The answer to question b is a tentative no, the differences at small sizes are not visible. I did not print this large (A3+ or larger) and I am not sure am going to. Perhaps another picture another time, which has more photographic merit to be printed large.

 

[Ben Rubinstein]

You shouldn't have had any problems with NPP on that image and with that lens, most of my stitching work has been without a pano rig, it's the wide angles (50mm or wider) and indoors where it starts becoming more necessary.

 

[Bart_van_der_Wolf]

You shouldn't have had any problems with NPP on that image and with that lens, most of my stitching work has been without a pano rig, it's the wide angles (50mm or wider) and indoors where it starts becoming more necessary.

Hi Ben,

You are correct that it is often possible to even stich images with some parallax. However, as you have found out the hard way, it can sometimes be very difficult (it even 'forced' you to buy another stitching software) and very time consuming if the software doesn't deal with the double contours from parallax well. And that is with short focal lengths. Cem's test involves longer focal lengths to increase the level of detail, and there the parallax errors become even more pronounced. Focus and DOF also become huge issues to reckon with when using long focal lengths.

Cheers,
Bart

 

[Ben Rubinstein]

I've never actually had a problem with focal lenths over 50mm outdoors with my project work. This first image was shot with a 70mm focal length, it's 38 megapixels worth of detail from eight 5D files. The next image was shot with a 100mm lens, it's 39.5 megapixels from 17 frames. Not a single problem or indeed with any of that series. The only problems I did have was when using a 50mm lens or indeed when I tried doing indoor work with a 16mm lens or a 8mm fisheye.

_

 

[Mike Shimwell]

Hi Cem

It makes me smile that you would shoot 135 frames of a wall to test whether stitching the 100mm was feasible compared to single frames from the 24mm The result is a clear increase in detail, but I wonder whether you are threatened with light changes during the long series of exposures required to shoot 132 frames (albeit only! 44 sets of braketed shots)? Still a good result and would likely look superb at 5ft wide and stil fine at 10ft wide (you don't really need to stand nose to nose with a 10ft wide print I think!

Asher, I'm not sure where you're going to find an output device that will put 100lpmm onto cibachrome from a digital file. Most seem to run about 300dpi, which equates to something 6lpmm

Mike

 

[Asher Kelman]

!

Asher, I'm not sure where you're going to find an output device that will put 100lpmm onto cibachrome from a digital file. Most seem to run about 300dpi, which equates to something 6lpmm

Mike

Are there laser printers that can use this fine capability of Cibachrome? In doing direct to paper exposure, the LF lenses might do 5-20 lp/mm at 50% contrast! So there's a lot of room to go with analog media if one would have the lenses to match!

But how, perceptually, might one better appreciate any better LF lenses?

Asher

 

[Cem_Usakligil]

Hi Mike,

..It makes me smile that you would shoot 135 frames of a wall to test whether stitching the 100mm was feasible compared to single frames from the 24mm ...

Sigh, am I that pathetic?
I should get a life, me!

Cheers mate...

 

[Doug Kerr]

Hi, Mike, Asher,

Asher, I'm not sure where you're going to find an output device that will put 100lpmm onto cibachrome from a digital file. Most seem to run about 300dpi, which equates to something 6lpmm

The Durst Lambda 131 HS will supposedly do 400 px/inch, which is about 8 lp/mm. I assume it will write to Ilfochrome.

Best regards,

Doug

 

[Asher Kelman]

Hi, Mike, Asher,



The Durst Lambda 131 HS will supposedly do 400 px/inch, which is about 8 lp/mm. I assume it will write to Ilfochrome.

For sure! But thats hardly using the full capability of the Cibachrome as a recording material.

Asher