Hey!
I've recently started using photoscan and i got to a point where all my models had the background trees or sky bleed into it, so naturally i tried to process a model with masking.
My results were nice, so i started to mask my next model as well.
I'm currently 3 hours into masking my next model because i have ~100 images, and according to my estimates it'll take 5 hours to finish making. So here i am, asking about masking because I'm trying to make sure that my work isn't going to waste.
What produces the best results while making? Specifically, where do you tend to mask the edges of objects? Here is an example of what I'm masking, and what i think the options are:
Unmasked:
http://puu.sh/lVkEb/03e4941df1.jpgMask edge accurately:
http://puu.sh/lVkLN/0d26e889ca.jpgMask edge accurately, mask out (fresnel?) reflection:
http://puu.sh/lVkNv/3e848bb4b4.jpgMask all edges out:
http://puu.sh/lVkQF/5ee9f1a1fc.jpgWhich of these would produce the best results? Because i figure for sparse cloud the program will still use the edges even if it's masked and the options are set that way, but you wouldn't necessarily want it to generate the model and textures based on the edge of the model, but rather when that edge part "rotates" into view.
In closing this is a pretty weird dilemma. The more images you take the better your result is, but the more images you have the more time you have to spend on masking, and it's not trivial at all. Most of the time you can't really pick or choose what your backdrop is, as 80% of the objects worth scanning are usually outside, in open environments.