A Segmented Consistency Check Approach to Disparity Map Refinement

Abstract

In stereo vision field, image processing techniques are used to increase the robustness of some types of solutions, especially those involved in real-world applications. With the objective to improve raw disparity maps, we investigated strategies of how to reduce the error measurement between a map and its correspondent ground truth. In this paper, we propose a simple but an effective technique, named as segment consistency check (SCC), to adjust a disparity map in a more appropriate configuration. The SCC method is organized into three main steps: segmentation process, statistical analysis, and by using adaptive support window. Experiment results show that the proposed method is comparable with other postprocessing strategies, in some cases with better results. Furthermore, through a qualitative evaluation, it is possible to note that the SCC method reaches significant results, especially in retaining object shapes, in maintaining the regularity of disparity in uniformed areas, and to preserve discontinuities.