Digital back stitching on a 4x5" analog system - is it feasible?

[torger]

I understand the fun of doing this, but is there any value in terms of results compared to spherical stitching that will generate a tack sharp image from corner to corner, will be much faster to shoot, will not have any color cast issues,...?

It is about the process. In one case you do spherical stitching, in the other case you digitize what you have composed in the view camera. If you don't see any value or enjoyment to work with the view camera then of course stitching with a modern 35 DSLR will be more natural, and that's all fine.

I don't think it will be much faster to shoot though, unless you skip tripod and non-parallax-point all together and shoot handheld. If you do that you must have good light conditions, and no foreground or not care about stitching errors. When I work with my click-stopped pano head it takes about 4 seconds frame-to-frame one part of the time is to wait for vibrations to subside (when I try to work faster I often end up with one image in the set being unsharp due to vibrations, *arrrghh*), with a bit more time to change row. I think it will take about the same time stitching at the back of a view camera, a fraction slower horizontal but faster vertical. In total it can actually be faster because you have well-defined borders already so don't need to shoot extra around. The current pano head I have you cannot adjust zero point vertically, it is always pointing straight at the horizon, which means that you must get quite some extra material on top/bottom if you don't want the horizon exactly in the middle in the final composition, that is either use a wider angle lens and get lower resolution or do more rows. If I'll continue use pano head stitching seriously I must fix that, I haven't found a manufacturer that has a fix for it, but nodal ninja whose ultimate edition I use now has kind of indicated that it will happen at some point (still waiting though).

If you look at what people generally do spherical stitching for it is either very wide panoramas 2:1 or 3:1 formats or wider of grand landscape views (top of a mountain or building or something, rarely with foreground) or spherical "VR" images. There's rarely high resolution images of "normal" compositions with normal aspect ratios (including portrait formats) with near foreground, or compositions that require tilt in the view camera (focus stacking if spherical stitching). Why? One could say that only those wide grand landscape views gain from high resolution, but I'd say most landscape images that has a distant horizon in it in one way or another can gain. The other reason is that it is much harder to stitch normal compositions -- you need tele lenses, no parallax point and deal with the DOF problem. Finding the corners in the composition becomes harder too, at the same time as composition becomes more precise and important to tune compared to a grand landscape view with no foreground.

[BernardLanguillier]

It is about the process. In one case you do spherical stitching, in the other case you digitize what you have composed in the view camera. If you don't see any value or enjoyment to work with the view camera then of course stitching with a modern 35 DSLR will be more natural, and that's all fine.

As I said, I understand the fun of using a view camera, but I am not sure how fun it can be to have to double each image with an LCC shot to be able to compensate for possible color issues...

I don't think it will be much faster to shoot though, unless you skip tripod and non-parallax-point all together and shoot handheld. If you do that you must have good light conditions, and no foreground or not care about stitching errors. When I work with my click-stopped pano head it takes about 4 seconds frame-to-frame one part of the time is to wait for vibrations to subside (when I try to work faster I often end up with one image in the set being unsharp due to vibrations, *arrrghh*), with a bit more time to change row. I think it will take about the same time stitching at the back of a view camera, a fraction slower horizontal but faster vertical. In total it can actually be faster because you have well-defined borders already so don't need to shoot extra around. The current pano head I have you cannot adjust zero point vertically, it is always pointing straight at the horizon, which means that you must get quite some extra material on top/bottom if you don't want the horizon exactly in the middle in the final composition, that is either use a wider angle lens and get lower resolution or do more rows. If I'll continue use pano head stitching seriously I must fix that, I haven't found a manufacturer that has a fix for it, but nodal ninja whose ultimate edition I use now has kind of indicated that it will happen at some point (still waiting though).

Really Right Stuff heads don't have this issue. For short shutter speeds I can do less than 2 secs per frame without any risk in terms of sharpness.

If you look at what people generally do spherical stitching for it is either very wide panoramas 2:1 or 3:1 formats or wider of grand landscape views (top of a mountain or building or something, rarely with foreground) or spherical "VR" images. There's rarely high resolution images of "normal" compositions with normal aspect ratios (including portrait formats) with near foreground, or compositions that require tilt in the view camera (focus stacking if spherical stitching). Why? One could say that only those wide grand landscape views gain from high resolution, but I'd say most landscape images that has a distant horizon in it in one way or another can gain. The other reason is that it is much harder to stitch normal compositions -- you need tele lenses, no parallax point and deal with the DOF problem. Finding the corners in the composition becomes harder too, at the same time as composition becomes more precise and important to tune compared to a grand landscape view with no foreground.

I do that a lot... DoF is no more a problem than with the equivalent MF camera unless you use movements of course. I don't see what's hard with normal compositions really, you just need to decide the corners of your composition. If you have a hard time vizualizing this you can just cut out a carton board with a typical 4x5 aspect ratio and move it back and forth in front of your eye to identify a suitable crop.



Many more after the link.

http://www.flickr.com/photos/bernardlanguillier/sets/72157600916381270/

Cheers,
Bernard

[torger]

Nice work you have there!

I'll see what I do, some more speed training with the pano head may solve some of the issues. I don't think there's any stability issues in the head itself (the NN ultimate is solid, gitzo tripod under it too), it is more that when working as fast as I can I mess up and hit the trigger almost before I let the hand off the camera or its ready in the clickstop, or accidentally hit some button on the camera. But I do need a better upper rotator, my current head has only 7,5 degree clickstops and no zero adjustment. 7,5 degrees is kind of okay, but without the zero adjustment it is limiting.

I would of course not get a view camera only to shoot digital. The most pleasing photographic process I have currently is working with my 24mm TS-E on the 5Dmk2, adjust tripod height and position, fine-adjust with the geared head, shift, tilt/focus, fine-adjust a bit more, wait for the moment, and take the shot. I'm a bit fascinated by those small adjustments, moving only a little bit can change quite much, it is more to composition than just identifying the corners, it is how foreground and background objects relate to each-other, how they overlap etc. This process is so much nicer than doing stitching that I find myself not using the pano head as much as I thought I would. Stitching at the back of a view camera as a complement to film would be more similar to that process. I actually do find composing with movements and seeing the picture in the view finder / ground glass / live view a nice experience in the photographic process.

As I said earlier, it is not only about the resulting file, it is about the process too, and there we all have different preferences of course.

(The LCC stuff is not worrying me personally too much any longer, my worry was if there would be significant DR loss after correction, but it does not seem like that. Now I see it more as an interesting problem, I do write some post-processing software hacks on my own (there's a reason I almost exclusively use open source software for post-processing), and the LCC problem seems interesting from that point of view...)