基于视差梯度约束的匹配点提纯算法  

Matched Points Purify Algorithm Based On Gradient of Disparity Constraint

作　　者：龚声蓉[1] 赵万金[1] 刘纯平[1] 

机构地区：[1]苏州大学计算机科学与技术学院,江苏苏州215006

出　　处：《系统仿真学报》2008年第S1期407-410,共4页Journal of System Simulation

基　　金：国家自然科学基金项目(60673092);教育部科研重点项目(207040);江苏省高校自然科学基金项目(07KJD520186)

摘　　要：在图像拼接过程中,粗匹配点对提纯算法RANSAC(Random Sample Consecsus,随机抽样一致性)存在计算量大,效率低等缺陷。为提高该算法提纯匹配点对的效率,基于预检验的策略,提出了基于视差梯度约束的RANSAC算法。实验结果表明,算法在不改变匹配精度的前提下,很好地提高了算法的效率。In the image mosaic process,the RANSAC(Random Sample Consecsus) algorithm of purifying rough matched points has large amount of calculation and low efficiency.In order to improve its efficiency,using pre-test strategy,this paper presents a RANSAC algorithm based on gradient of disparity constraint.The experimental results have shown that the new algorithm improves the efficiency effectively,and does not reduce the accuracy of the algorithm at the same time.

关 键 词：图像匹配 视差梯度约束 特征提取 RANSAC算法 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]