Dense disparity estimation based on the continuous wavelet transform [stereo image analysis]

This paper presents a new method for disparity estimation based on the continuous wavelet transform (CWT), by making use of the redundant information that results from the CWT. Several kinds of wavelets are tested for this purpose, including frequently used 1D wavelets as well as two nonseparable 2D orthogonal wavelets. The results are compared from the points of view of both the matching rate and preservation of feature information, and show that the disparity estimation based on the continuous wavelet transform can give better results than the standard area based methods SSD and SAD. In addition, a new method which uses zero insertion is introduced to increase the matching rate.