Software for a simple 2D panorama stitch from 3 PI-XAS scans

brbo said:

... and the rest, too. You may not care about that since the distortion is not huge, but some people would rather have the scan exactly the same as the negative. It's not any harder, you just need to use a proper tool.

Click to expand...

Your source images weren't entirely square-- Affinity recognized a triangular wedge between them, and corrected it. It's possible if I corrected for what looks like a very mild barrel distortion, and rotated the two images so that they were exactly parallel to each other, it would have produced a cleaner stitch.

In my stitched version, it's just barely visible as a dip in the negative frame on a vertical line slightly left of the "I" in FABRIS. If I truly cared, I could attempt to correct that distortion before stitching, but I preferred to take your images "as-is". Certainly my result was better than any of the examples you posted prior to this one.

So, yes, the frame is a bit wobbly, but the image itself, I can't find any blending errors or curvature of straight lines. Everyone has the requisite number of heads, arms and legs, so generally, for a zero-effort stitch, I'd say it came out pretty well.

Would you mind pointing out where the image itself is distorted in my version? I can provide a larger version if you like, Photrio just didn't want me uploading them (which is perfectly reasonable).

grat said:

Sorry-- implied was that if it can correct distortion, a non-distorted image should result in zero distortion.

Click to expand...

With what stitching workflow?

Radost said:

With what stitching workflow?

Click to expand...

Let's take Hugin or PTGui as an example. In both cases, you define congruent points on your images, the more the merrier. The software then (simplified explanation ahead!) analyzes the multiple points, and constructs a 3D model of a sphere, a cylinder, or just a plane, and projects your multiple images onto that surface, attempting to determine the best fit based on your points of alignment and (if I recall) the optical characteristics of the lens used (usually FoV is sufficient). It then "unwraps" that 3D object into a flat plane, typically creating what's called a rectilinear projection (imagine a flat map of the world, created from a globe).

So there's a whole lot of mapping and analysis going on. If your images all line up correctly, with no correction needed, then the final image should have no distortion in it.

I don't know what goes on under the hood with Adobe or Affinity, but I assume it's somewhat similar. As noted in my post above regarding brbo's images, because they weren't completely aligned (and may have a slight distortion in each image), the resulting image is slightly skewed, but only the black frame around the photo-- the image itself appears (at least to me) to be straight, with no obvious distortions in the image. A (slightly) cropped image would produce a rectangular image with no obvious stitching artifacts. The color is another issue, though. My inversion was terrible.

So, I've shown a .psd file where not a single pixel is out of place in the stitch and this is somehow worse than a stitch with a wavy frame.

Ok. I think I'm done here.

brbo said:

So, I've shown a .psd file where not a single pixel is out of place in the stitch and this is somehow worse than a stitch with a wavy frame.

Ok. I think I'm done here.

Click to expand...

Sorry-- I meant your examples that weren't inverted, specifically post #36.

I had forgotten the PSD.

The wavy frame is completely irrelevant. The contents of the image are more interesting. Again-- please show the distortions in the stitched IMAGE.