Artifacts along seams when relighitng

[yoterel]

Hello, thanks for this implementation !

I have ran the entire code on my own setup, but encountered a strange phenomenon of artifacts between pixels when relighting using patterns that are not "all 1's". This reproduced for 2 different scenes but does not happen with the provided data.

Have you seen such artifacts or have a way to counter this? any help would be appreciated.

Attached some photos to show what I mean (first image: synthetic image created from shining all 1's (works), 2nd,3rd: new synthetic lights that were used, 4th 5th: the synthetic image created by this implementation projecting these two lights). Notice the artifacts created where the light isn't shining.

[elerac]

Hello, @yoterel

Great implementation in your setup.

I'm sorry, but I don't know the exact reason for that weird artifact. I haven't encountered such a problem.

I guess you should

• Increase the number of projection patterns (1.5 to 2.0 times more than the minimum requirement). It can be robust against noise, and your artifact may be suppressed.

Additionally, make sure when you capture the images.

• Set the camera setting manually (i.e., exposure time, gain (iso), aperture, white balance). Do not rely on auto. It may change the setting while capturing sequential images.
• Disable the gamma correction. While estimating the light transport matrix, I assume the camera response function is linear.
• If you use a DMD or laser scanning-based projector, you must consider the camera's exposure time. Check the raw image, and it does not have a weird pattern.