Photogrammetry (3D Scanning) Diffuse Texture isn't "seamless" after Photoshop adjustments (Shadows/H

Jonathan85

Hello

Im doing a scan of smaller object (orange (fruit); but it doesnt matter, could be anything and the problem happened on and off previously with various scanned models). Im doing the diffuse texture in Agisoft Metashape from the original photos that were used for point cloud, mesh etc. generation ( i dont use another special, separate set of photos (adjusted) just for diffuse texture generation (as is often the case/recommended in a more proper photogrammetry workflow)).

I just use the Photoshop feature/function "Shadows/Highlights" (Image - Adjustments - Shadows/Highlights) to "somewhat" remove highlights and shadows from the Diffuse texture.

(I use indoor scanning setup on a rotating lazy susan with 2 softboxes for lighting so the lighting is "quite" (well... somewhat) even to begin with)


It usually works ok for me (for my "quality  level demand"), but SOMETIMES, as happened currently the (in photoshop) adjusted texture (if applied on the 3d object (in 3ds max)) suddenly gives/shows seams among the UV unwraped islands, picture:
https://prnt.sc/PVMLKlufpM0G

Original Diffuse texture (UNaltered in photshop in any way) doesnt have the seams:
https://prnt.sc/pMKC5ZcR6yKo

I dont know... I vaguely remember that it might have happened before sometimes (maybe..?)  but usually i dont see any problems... Now i do... My understanding is that the shadows/highlights function in Photoshop should work on the whole image "equally" so no "showing seams" problem should occur... Its like if your original diffuse texture doesnt show seams, and you for example, whatever... brighten, darken, add contrast, add color tint, change color balance etc. whatever you do with the texture in Photoshop, if there were no seams in the original diffuse texture, there should be NO seams visible even on the new adjusted one...? Right..? And it doesnt show AFAIK with brightening, darkening, changing color balance whatever...  BUT for some reason it shows (suddenly?) with Shadows Hightlights...? Why is that?

Again i think i MIGHT have gotten this problem MAYBE before but it was NOT often, i cannot even remember it really... But suddenly it shows with the newest model i work on... ( i didnt play with photogrammetry for months really)?

pior

Well, that's simply how Shadows/Highlights operates - basically working as an edge detection filter, similarly to Unsharp Mask.



Instead you need to use methods affecting all pixels the same : Levels, Brightness/Contrast, Curves, and so on.

myclay

What Pior wrote + some other tips;

• get out of the RGB mode.
  LAB works better in canceling out light information.
• you want a specular matte to erase that from the rest. To do that, you want to split the AOpponent and BOpponent from the Lightness (Lightmap).
  
• use Layer filters (Multiply, Linear Light, color etc.) to cancel the maps out over each other.
  

Jonathan85

myclay said:

What Pior wrote + some other tips;

• get out of the RGB mode.
  LAB works better in canceling out light information.
• you want a specular matte to erase that from the rest. To do that, you want to split the AOpponent and BOpponent from the Lightness (Lightmap).
  
• use Layer filters (Multiply, Linear Light, color etc.) to cancel the maps out over each other.
  

thanks, but i believe this is a bit out of my "pay grade" ("knowledge level" to be precise :-) ). Any link (to a tutorial or so) that explains what you are talking about in more detail...? Thank you :-)!

Jonathan85

pior said:

Well, that's simply how Shadows/Highlights operates - basically working as an edge detection filter, similarly to Unsharp Mask.



Instead you need to use methods affecting all pixels the same : Levels, Brightness/Contrast, Curves, and so on.

Ah, thanks... So i stand corrected, Shadows and Highlights do NOT affect the whole image equally... I did not know that :-)... And i used this method for majority of my few hundred scans lol :-)... Only rarely i noticed any seam problems... ehm... :-) was lucky probably

Also - Do i understand it correct, that the proper workflow for this would be to make a separate set of original photos (from RAW to TIF) JUST for the diffuse texture generation in Metashape...? This 2nd set of photos (JUST for texture generation, not point cloud, mesh etc., generation, JUST for the texture) should have shadows and highlights reduced even more (from the RAW photos)...? That means reduce the highlights and shadows on PHOTOS (via lightroom for example or Adobe Camera Raw) from which you generate the texture, NOT on the texture itself...?) Correct?

myclay

Jonathan85 said:

thanks, but i believe this is a bit out of my "pay grade" ("knowledge level" to be precise :-) ). Any link (to a tutorial or so) that explains what you are talking about in more detail...? Thank you :-)!

For the RGB to LAB conversion,look up LAB + Program you use.
I use each channel data and copy that into Layers.


Copy the contents of the LAB channels into Layers and lay them ontop of each other.


Disclaimer; this is more of an averaging of the Shadow and Light information within the A and B channels.
With additional HSL and Vibrance Adjustments it can be changed further.

EDIT; if you want to shoot images yourself, you could look into Cross Polarization.
https://youtu.be/yhjKO1a99OQ?t=21

Edit2:
https://www.pix-pro.com/blog/post/cross-polarization