Stereo Correspondence Based on Block Segmentation and Entropy Selected Optimization
Stereo Matching, Entropy, Block Correlation, Dynamic Programming
The stereo matching process presents several solution proposals. In each of them, there is a dilemma involving the execution time and a quality of the results obtained, usually on a disparity map. This project proposes a stereo matching algorithm in which as images are divided into blocks, and each block is processed independently by a technique of an available set. A choice of technique and calculated entropy valuation report for that region. Preliminary experiments are available to analyze some techniques available in the literature and to verify the use of entropy as a measure of texture. An implementation of the proposal and local and semi-global technique, and the results are compared with individual implementations of each technique. It is noticed that there is a reduction of the execution time, but problems are verified at the junction of the maps.