基于两级扫描线优化的快速立体匹配算法

Fast Stereo Matching Algorithm Using Two-Pass Scanline Optimization

通过引入视差梯度理论,提出了一种弱连续性约束,它表示相邻点的视差变化总在一定的范围之内。同时引入了一种两级扫描线优化方法,它从水平和垂直两方向进行扫描线优化。最后把弱连续性约束与两级扫描线优化方法结合起来得到一种快速立体匹配方法。通过对比实验表明,与未加入弱连续性约束的两级扫描线优化算法相比,所提算法计算量更少,同时减轻了同纹理区域的匹配不确定性;与传统一维扫描线优化算法相比,该算法具有更高的匹配准确性并解决了扫描线间的不连续性。

Based on disparity gradient theory, a weak consistency constraint was proposed, which expressed that the disparity difference between neighbor points was always in certain range. Meanwhile, a two-pass scanline optimization method was present, which processed optimization along and across the scanlines. At last, the weak consistency constraint and the two-pass scanline optimization were combined in a fast stereo matching algorithm. The experiment results indicate that the proposed algorithm has less computational cost than the two-pass scanline optimization algorithm without weak consistency constraint, and lessens the matching ambiguities in the similar textured regions. The algorithm is more accurate than the conventional single-pass scanline optimization algorithm and resolves inconsistency between scanlines.