Improving my stitching workflow

[GordonShock]

Hello all,

I primarily engage in photo restoration work, often dealing with the task of stitching together photos that are divided across two separate sheets. While I have accumulated some experience in this area, I recognize that there is room for improvement in my current technique. Allow me to outline my existing workflow:

All my photos are managed within LR, primarily for cataloging purposes.

When processing a photo, it is sent to PS, where one of two scenarios unfolds. If stitching is not required, I proceed to crop, descreen, restore, save, and then move on to color grading in LS.

However, when stitching is necessary, I send both image files to PS. In this case, I begin by cropping and erasing as much unnecessary padding from both photos as possible.

Subsequently, I double the width of the canvas, reposition the first photo to the left, and seamlessly integrate the second part alongside it. This step involves ensuring both parts are scanned with the same height and correctly aligned to save time.

At this juncture, I encounter two common challenges. Firstly, even though the files share identical dimensions and positioning, they often fail to align perfectly due to perspective discrepancies. My question is whether it is advisable to correct this perspective issue in LS or PS? If the latter is recommended, I would greatly appreciate guidance on using perspective correction tools effectively, as I have yet to explore this aspect.

Once I achieve a satisfactory stitch, the next task is to remove the stitching artifact. At this point, I merge both layers and attempt to use the Remove tool, but unfortunately, it proves ineffective. I have repeatedly tried to utilize it, but it consistently fails to remove the stitch as expected. The generative tool is also somewhat finicky, and while the spot healing tool works well, it may not be the most suitable tool for this particular task. My proficiency with the clone stamp tool has improved, but its applicability remains case-dependent; there are instances where it cannot be employed effectively, yet for other stitches, it suffices.

I am eager to enhance my workflow. Any advice or suggestions on how to streamline and improve this process would be greatly appreciated.

[Conrad_C]

When you use the Photomerge dialog box in Photoshop, which Layout option do you use? For flat art, the last two options, Collage or Reposition, are recommended. The other options assume you took photos of a real scene and need to apply a projection (such as Spherical) to rebuild the real-world perspective, but Collage and Reposition do not do that because they’re intended for reassembling flat documents. With Auto, you don’t know what you’ll get. So choosing one of the last two is recommended for flat art. Collage will try to correct things like one scan being rotated more than another, and I think Reposition only moves the scans to align them, with no other adjustments.

If you need to correct perspective adjustments, Photoshop has a number of perspective tools for different photographic situations, but for a flat scan it’s probably best to choose Edit > Free Transform, then hold down the Ctrl key (Command key on macOS) while dragging a corner handle to correct rectilinear distortion.

[GordonShock]

Thank you Conrad. I gave the Photomerge multiple shots, and all I got from my three attempts to merge a total of 3 pairs of scanned images was the same error message that they couldn't be merged. So I'm back to doing it manually. 

Most of the scanned photos I work on seem to have a barrel distortion, not a rectilinear one. It's extremely subtle, and I don't bother with, but when it comes to merging two parts of one photo, the slightest distortion affects the cropping and makes merging a lot more time-consuming.

My workflow starts with LS, but cropping in LS is overly clunky and you can't zoom in, which ensures that you'll end up cropping a whole lot more pixels than you should.

I'll give the "Lens Correction" a try, as well as your free transform suggestion.

Thanks

[D Fosse]

The Auto-Align command, which is a basic component in the composite Photomerge script, does not work without solid overlap. It doesn't look for fitting edges; it looks for pattern overlap. Around 40% is required.

By the sound of this, it sounds like a manual job. Convert to smart object to avoid cumulative degradation. I don't know of any automatic procedure that can work to fit edges without any material overlap.

[Conrad_C]

Photomerge should be able to deal with a modest amount of distortion, because it can merge real world panoramas shot with wide angle lenses. But if you want to try and remove it as a variable, then you can try using the Distortion option in Filter > Lens Correction, Manual tab to remove the distortion from each image before attempting a merge. If they were originally shot in raw format, you can also try the Distortion option in Adobe Camera Raw or Lightroom Classic, in the Optics/Lens Corrections panel, Manual tab. But if the barrel distortion is as subtle as you say, it’s probably not the problem.

If any of the samples are OK to show publicly, it might help to add them to this thread and we can take a look.