Semester Master Project
Microengineering section, EPFL
In this work, our intent is to do stereo-matching using pyramidal complex wavelet transform. We will present a discrete complex wavelet transform based on Gabor-like wavelets and see that it exhibits a good translation invariance, which insures that the spacial disparity will be encoded in our transform. The translation invariance is especially better for the transform we present than the standard ones, using only real wavelets. This last property will be used to build a pyramidal patch-based \coarse-to-ne" optimization, that reduces the computation time. We will present such algorithm, talk about the issues related to this one and nally show some results.