This paper has proposed a phase matching algorithm based on wavelet transform. This algorithm first uses the classical wavelet transform-- Mallat pyramid method to decompose the original images into bandpass of different scales, and through the Hilbert transform to convert it into the corresponding analytic bandpass signal, and then follow the usual phase matching method to conduct disparity estimation from the order of coarse to fine. Using this pyramid structure of multi-scale matching, we can significantly improve the matching efficiency. In addition, targeted at the fact that phase winding appears in certain areas, this algorithm has achieved adaptive interpolation of these regions by using the smoothness constraints of disparity in the matching process, and therefore improve the stability of this algorithm.
展开▼
