A New Disparity Selection Method Based on Extreme-Point Extraction

Abstract

Disparity selection is an important step in stereo matching and the Winner Takes All (WTA) strategy is commonly used to compute the disparity map from the aggregated cost volume. However, this strategy is lack of robustness, a contradiction between the accuracy of the result and the complexity of the method would be caused and the application of the local method is limited severely. In this paper, the defect of the WTA strategy is elucidated by introducing the concept of the cost curve, firstly. And then, we present a new disparity selection method including two steps. The first step, some initial points are extracted from the cost curve. The second step, the matching point would be found by a discriminator in the initial points. Finally, a complete local matching framework is established and the standard image data sets are used to verify the effectiveness of our method. The experimental results show that the proposed method performs much better compared with the WTA strategy and produces good results, even though a simple cost aggregation is used.