A new global energy optimization framework for stereo disparity estimation

Abstract

Stereo matching is an essential step in depth estimation, which can be formulated as an energy minimization problem. This paper presents a novel global energy optimization framework for stereo matching. We formulate a new data term revised from normal NCC algorithm. In addition, we adopt multichannel pixel values from input corresponding images to improve the data accuracy. Finally, a bidirectional stereo matching strategy is introduced to refine the disparity calculating results. Our proposed method is definitely helpful for stereo matching and reconstruction of stereo scene. Moreover, we have compared the proposed method with both local method and the approach in [4] quantitatively and qualitatively. The results based on Middlebury data sets show that our method outperforms the others apparently.