A hierarchical block matching stereo algorithm based on cepstral analysis for remote sensing
Juan M. Ramírez, Otto J. Paz-Luna, Heber Salazar · 2008
This paper presents a stereo vision system based on a hierarchical matching process using cepstral analysis in the disparity detection process. The use of the Cepstrum transform as a measure of matching support allows the automatic decision of the block size during the analysis of the stereo images. A fast Hartley transform algorithm is incorporated for execution time improvement. The automatic determination of local similarity between two data sets or collections of pixels is fundamental in a depth-from-stereo algorithm for numerical evaluation of disparities. This measurement process is embedded within a coarse to fine stereopsis algorithm, providing an initial depth map with the depth information encoded as gray levels. The proposed algorithm has been tested using random dot stereograms and remote sensing imagery stereo pairs. Index Terms—cepstrum, images, matching, stereo.