基于前向搜索的快速视差范围估计方法
发布时间：2019-05-07 20:37
【摘要】：在立体匹配中,设置合理的视差搜索范围能够提高匹配的速度和精度。为此,提出了一种基于前向搜索的图像迭代细分方法用以估算视差范围的上下限。将参考图像均分为若干个图像块,在对每一块的匹配过程中,采用前向搜索策略,对当前匹配块继续均分成若干子块,并对其子块进行相似度匹配。在计算视差范围上限时,用当前块的子块视差中的最大值来表示其视差值,找到视差最大的图像块后继续迭代细分,直到得到稳定的视差范围上限。在计算视差范围下限时,用当前块的子块视差中的最小值来表示其视差值,找到视差最小的图像块后继续迭代细分,直到得到稳定的视差范围下限。实验结果表明,采用前向搜索策略计算视差范围的方法,在搜索空间降低比率达到28.8%的同时能够达到98%的命中率,相较直接进行匹配误匹配率降低了47.4%,能够得到更精确的视差范围。
[Abstract]:In stereo matching, the matching speed and precision can be improved by setting reasonable disparity search range. Therefore, an iterative image subdivision method based on forward search is proposed to estimate the upper and lower bounds of parallax range. The reference image is divided into several image blocks. In the matching process of each block, the forward search strategy is used to divide the current matching blocks into several sub-blocks, and the similarity matching of the current matching blocks is carried out. When calculating the upper limit of the parallax range, the maximum value of the sub-block parallax of the current block is used to represent its parallax value. After finding the image block with the largest parallax, we continue to iterate the subdivision until a stable upper limit of the parallax range is obtained. When calculating the lower limit of parallax range, the minimum value of the sub-block parallax of the current block is used to represent its parallax value. After finding the image block with the smallest parallax, the image block with minimum parallax is further subdivided iteratively until a stable lower limit of parallax range is obtained. The experimental results show that using the forward search strategy to calculate the disparity range, the search space decreases by 28.8% and the hit rate reaches 98%, which is 47.4% lower than the direct matching error matching rate. A more accurate range of parallax can be obtained.
【作者单位】： 东北林业大学信息与计算机工程学院;
【基金】：国家自然科学基金(No.31370565,No.61300098)
【分类号】：TP391.41
[Abstract]:In stereo matching, the matching speed and precision can be improved by setting reasonable disparity search range. Therefore, an iterative image subdivision method based on forward search is proposed to estimate the upper and lower bounds of parallax range. The reference image is divided into several image blocks. In the matching process of each block, the forward search strategy is used to divide the current matching blocks into several sub-blocks, and the similarity matching of the current matching blocks is carried out. When calculating the upper limit of the parallax range, the maximum value of the sub-block parallax of the current block is used to represent its parallax value. After finding the image block with the largest parallax, we continue to iterate the subdivision until a stable upper limit of the parallax range is obtained. When calculating the lower limit of parallax range, the minimum value of the sub-block parallax of the current block is used to represent its parallax value. After finding the image block with the smallest parallax, the image block with minimum parallax is further subdivided iteratively until a stable lower limit of parallax range is obtained. The experimental results show that using the forward search strategy to calculate the disparity range, the search space decreases by 28.8% and the hit rate reaches 98%, which is 47.4% lower than the direct matching error matching rate. A more accurate range of parallax can be obtained.
【作者单位】： 东北林业大学信息与计算机工程学院;
【基金】：国家自然科学基金(No.31370565,No.61300098)
【分类号】：TP391.41
【相似文献】
相关期刊论文 前10条
1 刘洪江;李丽;;基于稀疏测量的稠密视差图重构研究[J];价值工程;2012年18期
2 贾波,章毓晋,林行刚;视差图误差检测与校正的通用快速算法[J];清华大学学报(自然科学版);2000年01期
3 王爱红;王琼华;李大海;李小方;;会聚式相机获取的视差图像的梯形失真校正[J];四川大学学报(工程科学版);2011年02期
4 宁静静;孔令德;;一种基于双向互匹配的视差图算法[J];电脑开发与应用;2012年01期
5 孔庆群;高伟;;视差计算的层级模型[J];中国科学:信息科学;2013年09期
6 赵戈;蔺，
本文编号：2471364
本文链接：https://www.wllwen.com/kejilunwen/ruanjiangongchenglunwen/2471364.html
