一种室外非理想光照条件下的立体匹配算法

针对室外非理想光照条件和图像低纹理、低对比度造成立体匹配效果较差的问题,提出一种HSL(色相-饱和度-亮度)颜色空间下基于边缘图分割的立体匹配算法.区别于传统的RGB颜色空间下基于像素强度的度量方式,该算法采用一种HSL颜色空间下的像素非相似性度量公式来获得匹配代价,然后基于左右输入图像的边缘检测结果进行图像区域分割和立体匹配.在实验中采用一系列不同光照条件的图片集和具有明显低纹理区域的图片集,对本文算法与现有算法进行对比验证.实验结果证明,该算法能够得到比较理想的视差图,对非理想的光照条件和低纹理图像具有很好的鲁棒性,并且基本达到实时性要求.

A Stereo Matching Algorithm under Outdoor Unideal Illumination Condition

To solve the problem of poor stereo matching effect caused by unideal outdoor illumination conditions and low-texture or low-contrast areas,a new stereo matching algorithm is proposed based on edge-map guided segmentation method under HSL(Hue-Saturation-Luminosity) color space.Different from traditional pixel-intensity based measurement under RGB color space,an HSL-color-space based pixel dissimilarity measurement equation is used in the proposed algorithm to calculate matching costs,and then image region segmentation and stereo matching are completed based on the edge detection results of the left and right input images.A series of image sets with different illumination conditions and significant low-texture areas are used for comparison between the proposed algorithm and the existing methods.Experiment results show that the proposed algorithm can achieve ideal disparity maps,be robust for images with unideal illumination and low-texture areas,and run in near real-time.