摘要
提出了基于区域的立体匹配算法,实现从彩色立体图像对中提取深度信息。其中利用平滑函数求取全局错误能量最小化获得的视差图更为可靠,但是计算时间也较长;而采用线性生长算法获得的视差图则计算时间短,但可靠性差。结果是利用滤波去掉了不可靠的视差估计,可以提高视差图的可靠度,还对算法生成的结果进行了比较。
In the paper,region based stereo matching algorithms are developed for extracting depth information from two color stereo image pair.It shows that global error energy minimization by smoothing functions method is more reliable but more time consuming,and that line growing method is more convenient for the sequential computing architectures because of promising higher speed.Filtering unreliable disparity estimation can increase reliability of disparity map.Finally,a comparison of the algorithms is presented.
出处
《南京邮电大学学报(自然科学版)》
2011年第3期44-49,共6页
Journal of Nanjing University of Posts and Telecommunications:Natural Science Edition
基金
国家自然科学基金(61071091)
国家自然科学青年科学基金(61001152)
南京邮电大学校科研基金(NY210053
NY210069)资助项目
关键词
立体匹配
视差图
块匹配
线性生长
stereo matching
disparity map
block-matching
line growing