Metashape chooses blurred patches for orthomosaics

[NMS]

We are using Agisoft Metashape 2.2.2 on Windows 11 for documenting commingled human bones on an archaeological dig. A typical dataset of ours comprises about 100 photos shot by hand for an area of about 4–5 square metres. Because of the human bone chaos, the photos need to be taken from very different angles, so not every part is sharp in every photo, of course, but in the Metashape quality check the photos show a value of 0.69-0.83. The workflow runs smoothly with default settings, and usually 100% of our photos are aligned properly.
Strangely, in many patches of the resulting orthomosaic, not the best photo is chosen but a very blurry one. Of course, this can be corrected by hand by selecting a different set of photos for the incriminating patch, but this manual work is time-consuming. It also seems unnecessary because even the second and third photos of this patch are much sharper and therefore better suited.
To make things even more irritating, the same workflow, set of photos, and even cameras were successfully used in past years without such problems. Therefore, I do not understand why Metashape chooses blurry photos with the current version. We have already consulted the manual, played around with the settings, and searched this forum but have not found any obvious answers. I appreciate any hints how to solve our problem.

[Alexey Pasumansky]

Hello NMS,

Metashape takes into account several criteria when choosing optimal images for orthomosaic. Those include distance from camera to the area of interest, effective image resolution for the area of interest, how close is camera normal to the normal to surface in the area of interest and some others.

If you have some blurry and out-of-focus areas on the source images, you can try to generate masks for them in automatic mode using Generate Masks command, from Model method and with "mask defocus areas" option enabled. After that re-build the orthomosaic confirming that unwanted areas are properly masked.

[DayGeckoArt]

Hello NMS,

Metashape takes into account several criteria when choosing optimal images for orthomosaic. Those include distance from camera to the area of interest, effective image resolution for the area of interest, how close is camera normal to the normal to surface in the area of interest and some others.

If you have some blurry and out-of-focus areas on the source images, you can try to generate masks for them in automatic mode using Generate Masks command, from Model method and with "mask defocus areas" option enabled. After that re-build the orthomosaic confirming that unwanted areas are properly masked.

I've had the same issue as the OP, I probably first observed it 3 years ago and posted about it. The solution was manually patching.

But what I noticed is often I would patch and the top ranked image would be a much better one. So I believe there's a bug. The program knows a sharper image is best, but doesn't choose it the first time.

I don't know how it works but it seems like it should be simple for the software to prioritize nadir images when generating an orthomosaic, and more difficult for software to evaluate sharpness. But really I think this is just a bug not a flaw in how the program works