It would take a lot of time to fully assess the capabilities of the various stitching programs available.
Two main issue for me are, (1) How well can the stitching program do a fully automatic stitch, because a perfect automatic stitch clearly saves a lot of time? (2) How well can the program handle the serious parallax errors that occur with close-up objects in the scene?
I was always aware that Panavue's Image Assembler was not too good at correcting for serious parallax errors in close-ups, which is why I ended up buying a TS-E 24mm.
However, I'm now getting the impression that both CS3 Photomerge and Autopano Pro can do a near perfect job with such images and that Image Assembler has got left behind in this regard.
The following shots were taken whilst trekking between Jomsom and Kagbeni, somewhere in the middle of Nepal. I was in awe of the sheer vastness and barrenness of this landscape. My 15mm Sigma lens on my Canon 5D was simply not wide enough. I needed to stitch a few images. I had my TS-E 24mm at hand. I wasn't sure if 2 or 3 stitched images with that lens would be wide enough but in any case my tripod was with the porter who, as always, seemed to race ahead despite his heavy load.
I wondered if I should send my guide, who always kept close to me, to retrieve the tripod. But the wind was beginning to blow in increasingly heavy gusts. This was a gorge which, regular as clockwork, developed strong winds around 10-11am each day. Strong enough to prevent the planes from landing.
I decided it would be a waste of time using a tripod in these conditions and took a series of 4 handheld shots at 15mm and f16, with big overlaps. I was concerned about getting the pebbles in the immediate foreground sharp. I wanted the maximum DoF my system could provide.
In restrospect, I think I probably miscalculated. I'd not taken into account the fact that the closest parts of the foreground would be cropped off after stitching. I think I could have focussed just a little further afield and perhaps (not sure though) got a slightly sharper distant horizon. I should have done some focussing bracketing.
Maybe I didn't think the scene warranted such attention, from a compositional point of view. However, those pebbles in the immediate foreground are literally almost at my feet. This is surely an ideal set of images to test these stitching programs.
Below are the results, but some explanation first.
(1) All images appear to have perfect joins, except with regard to tonal transitions in the sky. Image Assembler shows a slight darkening of the sky in the transition between first and second images.
(2) The distortion of the over all image is unacceptable in the Image Assembler stitch, but that was the best I could do in the time, meticulously placing 5 pairs of flags at each join.
IA simply couldn't manage a fully automatic stitch.
(3) All the CS3 Photomerge stitches are literally fully automatic. Just load the images and click OK. No prior settings to be made.
(4) The Autopano Pro stitch could be described as semi-automatic. One really has to examine the settings menu first, before stitching, and change the default settings to something one imagines might be more appropriate.
(5) The Autopano stitch uses more interpolation, producing the largest file.
In this example, Autopano 199.7MB; IA 168.8MB and CS3 just 139.2MB.
Immediately below are the uncropped resulting stitches from IA, Austopano and CS3.
I was interested if the significantly larger file that Autopano produced provided any more detail than the CS3 stitch. Very marginal I think, and not an issue.
I made a 100% crop of the pebbles near my feet in the area of the join between the first and second image. No problems at all.
I'm imprerssed with the progress that certain stitching programs have achieved. For static subjects, I think MFDB might be on the verge of redundancy .