Investigation of Using the Local Directional Pattern for stereo corresponding problem

Despite the significant success of using local feature descriptor with many applications such as, face detection, human facial expression recognition, gender classification, signature verification, etc. Based on our knowledge, there is no research present on the usage of the descriptor to solve the stereo corresponding problem. Despite this, descriptor uses the same methodology for comparing the images blocks. In this paper, we developed a new method to solve the stereo vision corresponding problems by using the Local Directional Pattern (LDP). The proposed method encodes the input images using LDP encoder, then for generating depth map, we use block-matching algorithm but we change distance measurement (chi-square) instead of similarity measurements to compare the blocks code histogram. The proposed method shows promising results comparing with other methods on testing images benchmark.