Surface normalization algorithm
I have a surface(in 3D space) which is an approximation of a human face, based on a 2D photograph(I use the brightness to extract depth).
The majority of the points have an acceptable depth(I only change Z-values since X,Y are corresponsive to the photo).
Since it is an approximation, there are many deviations from acceptable values and in many cases, the depth assigned is completely wrong. However, the mistake is ,somehow, local. I mean in most cases,the point or the collection of points with the wrong depth are surrounded by an area with acceptably correct value.
For example, a spot in the face results in a great depth deviation from the surrounding face.
My questions are:
1) How do I "normalize" the surface(bring the spot at the depth level of its surrounding surface)?
2)What is the name of this process(if not normalization), so as to search the web for further algorithms.
I thank you.
If the artifacts (deviations) are very local, then use simple median filtering. Basicly, take a 3x3 window of depth values around the current depth value and find the median of those 9 values. Replace the current value with the median. You can obviously use different window size or shape. This is a well known method in image processing. Most often it is used to remove impulse noise (or salt'n'pepper noise). Some info about median filtering
If you google mesh smoothing or denoising there is a lot of literture on this:
http://mrl.nyu.edu/publications/denoise/
http://www.mpi-sb.mpg.de/~shin/Research/NL/NonLocal.html
http://mrl.nyu.edu/publications/denoise/
http://www.mpi-sb.mpg.de/~shin/Research/NL/NonLocal.html
This topic is closed to new replies.
Advertisement
Popular Topics
Advertisement