The discussion about alignment reminded me of another question concerning masks.
I think somebody has asked a similar question somewhere, but I can’t find it right now.
Seems to me that you don’t want to introduce masks for the cameras, which I do not mind per se because your software seems to be doing an amazing job with difficult shots (e.g. sky or deep shadows). However, I am somewhat concerned about changing features in images like people, moving branches, birds and such. How does the software deal with these difficulties?

Hi Götz

We can use masking in other parts or the process, but for now its not implemented for Alignment step, we have some technological reason for that (not that we cannot do it )
For the question regarding changing features and how we handle it, the algos responsible for alignment take in account this, but yes just to some degree of differences. We planing to do some changes to the alignment. Stay tuned…

Hi Wishgranter,

ok, no problem (I hope :slight_smile:
Don’t worry, I will!

Here is the other discussion on masks >

Turntable shooting scenario


To explain the masking in RealityCapture - even we do not have a tool for editing masks build in our app, the app itself supports masks to some extent. For alignment step all you need to do is to modify image color channel and paint it with a solid color. This will cause that natural features will not be detected there and thus will not influence camera alignment. For meshing we support alpha-channel masks. Adding alpha channel and masking only important parts actually speeds up whole computation. Another benefit is that you would not need to use reconstruction region to filter parts which are not important.

You should be able to generate these masks with 3rd party software in batch process easily.

Hi martinb,

thanks, that was it!
I don’t quite understand the process though, could you describe it a bit more detailed?
Which channel do I use? Would alpha work for both, alignment and meshing?
Not sure if I have software (photoshop elements 12) that can save into alpha though - can you reccomend some freeware?

Hi Götz

just say have a house with backgroud stuff ( trees or other houses )

just select and paint everything except the house ( main interest area ) with single color say white…

and its like you used masking as on the plain white color no features will be detected

Hi Wishgranter,

Sounds good since it’s much more straightforward than all the alpha stuff… :slight_smile:
How would it work if I wanted to change a mask of an image - could I just replace the file and restart RC and it would load the updated version or would I need to realign?

Hi Götz

No need to reload imgs, if the files are changed, just clear cache and hit align again. so the points are detected from scratch.

Out of curiosity, is it currently possible to use a clean plate/difference matte?

In my case, I have an array of cameras that are fixed in position, angle and focal distance. It would be very useful for automatic masking to be able to provide a set of images that have just the stage and background, and then a second set with the actual subject.

Obviously, I could create an alpha mask by difference key in a compositing application, but it would be very useful to have this built into the software as it would remove additional steps.

Hi Neil.
if you use it in Fullbody scaner then there is no need to create masking.
have some specific issues that need to solve with masking ?

No, no specific problem, not yet at least. Just idle curiosity.