2BEDM: Brightness-Based Enhancement for Disparity Maps

Stereo vision is a technique used to recover depth information from stereo (two) images that capture the view from different perspectives. The stereo-matching/stereo-correspondence process presents an essential role in stereo vision. Much research has been done on this process. However, the accuracy of stereo-matching/stereo-correspondence algorithms is affected by many characteristics related to the scene, such as the lack of discriminative image features, repetitive patterns, specular reflections, and regions of uniform texture. In this paper, we propose a pre-processing schema called brightness-based enhancement for disparity maps (2BEDM). 2BEDM is used to enhance the accuracy of local stereo-matching/stereo-correspondence algorithms (which are affected by the aforementioned characteristics related to the scene in stereo images) and to enhance the output-disparity map of stereo-matching/stereo-correspondence algorithms. It is based on the brightness of each pixel in the stereo images (left image and right image) rather than on the pixel itself. 2BEDM works on the basis of the hypothesis that the brightness of each block of stereo images of the same views is equal. The experimental results illustrate the enhancement provided by the proposed 2BEDM schema.