融合颜色不变量的彩色图像光流估计算法

摘要

To overcome the influence of illumination and improve the accuracy of optical flow estimation for color images, a new optical flow estimation algorithm is presented, which fuses optical flow calculated in the color invariant space and that in the RGB space. First, the color invariant edges of the two adjacent frames in the video sequence are extracted. Then, the three-channel color images are built with the color invariant edges, and the optical flow of the color images is computed. Finally, the computed optical flow of color invariants is fused with that of the RGB space based on the L∞ norm. How the fused optical flow estimation can be influenced by different selection and combination of color invariant edges are also discussed. Experimental results show that the proposed method captures object areas more accurately, and is more robust to the shadow and illumination variation than the typical algorithms.%为了克服彩色图像光流估计中的光照影响，提高光流估计的准确性，该文提出一种在颜色不变量空间求解光流，再与RGB空间光流进行融合的光流估计算法。首先提取视频序列中相邻两帧图像的颜色不变量边缘，再计算使用颜色不变量边缘构建的三通道彩色图像的光流，最后使用L∞范数融合颜色不变量光流与RGB图像光流。文中还讨论了颜色不变量边缘的选取和组合对融合光流结果的影响。实验结果表明，提出的方法能够获取更加准确的目标区域，对阴影和光照变化更加鲁棒，与经典算法相比具有更好的性能。