Lately I have been doing some facial scans and I have noticed that my alignment is far superior if I desaturate the images before calculation, and my mesh accuracy is greater than if i used the color imagery.
Sadly all of the images I used are under NDA so I have no current examples I can send through, however having this as an option in-software would be great!
Hi erikjameschristensen
Would not it be better to have an additional image set where you do the changes with an external app that is better prepared to make the image adjustment with all the necessary fine-tuned tools ??
Hi,
I actually found something very similar when working with raw data. I shoot in raw, batch - process in photoshop where I basically bring highlights down and shadows up - this makes a far better result…Didn’t test desaturating yet.
Although there are probably better apps to prepare such data, there is no clue for the user what is the best possible color correction, so this is very much about very long experimentation. If there would be any clue at what color ranges are best for the algorithm, it would make our work much easier…
Hi vilem.duha
It‘s just about good images overall, no special settings / adjustments are needed in most cases. Just use a good RAW workflow ( DXO Optics ) and proper capture for photogrammetry with a lot of overlap…
Thanks, maybe using Photoshop is old school
Wil try DxO.
Thing is, I am building a human-scanner, which means I am limited with the number of cameras ~ overlap… So squeezing as much as possible from relatively few cameras is crucial for me. So far getting better and better results when experimenting with raw in combination with noise projection …