基于自适应窗的立体相机视差图优化方法研究

Research on Stereo Camera Disparity Map Optimization Method Based on Adaptive Window

针对通过立体匹配直接得到视差图存在空洞等失真点(区域)的问题,本文提出一种基于自适应窗的视差优化方法.该方法先通过左右图像的视差一致性,检测出非零视差失真和黑洞失真区域,然后将所在像素点与周围像素点作视差代价比较,并针对由环境、光照以及其他噪声因素造成的误匹配所导致的黑洞区域,通过采用自适应窗填充黑洞失真区域,进而利用中值滤波平滑图像,最终实现对由立体相机左右图像匹配所获得的视差图的优化.通过与3种现有的视差优化方法进行对比,结果表明:本文方法具有更好的视差优化效果,且实现更为简便.

Aiming at the problem that there are cavity and other distortion points(area)in disparity map directly obtained by stereo matching, this paper proposes a disparity optimization method based on adaptive window.Firstly, the non-zero disparity distortion and cavity distortion area are detected by the disparity consistency of left and right images and the disparity cost of the pixels is compared with the surrounding pixels.For the mismatching-conducted cavity from the environment, illumination and other noise issues, the distortion area of cavity is filled by using the adaptive window as well as the image is smoothed by median filter further.Finally, the disparity obtained by matching the left and right image from the stereo camera is optimized.Compared with the three available disparity optimization approaches, the proposed method shows the better performance and the convenience of implementation.