行列双动态规划的改进自适应立体匹配算法  

Improved adaptive stereo matching algorithm with double dynamic programming with row and column

作　　者：赵明[1] 李晓白[1] 郎荣玲[1] 

机构地区：[1]北京航空航天大学电子信息工程学院,北京100191

出　　处：《计算机工程与应用》2011年第23期177-180,共4页Computer Engineering and Applications

摘　　要：在各种立体匹配算法中,利用动态规划算法求解可有效地提高立体匹配的速度和精确度,同时具有实时性好、易于实现的优点。利用动态规划算法的优点,提出一种基于行列动态规划的自适应立体匹配算法,采用改进的自适应代价函数和能量最小化模型,对最优化问题进行求解。在求解的过程中,基于行动态规划得到的列方向视差值的变化给予对应数据项不同的奖励值,以减少行动态规划产生的明显条纹,最后使用列动态规划得出最终结果。实验结果表明,该算法能够减少总体的匹配错误率,减少明显的条纹瑕疵,取得较理想的立体匹配效果。Among various stereo matching algorithms,the one with dynamic programming can effectively improve speed and precision with good real-time performance and easy realization.This paper,based on advantages of dynamic programming algorithm,presents a self-adaptive stereo matching algorithm using dynamic programming with row and column,which uses improved adaptive support-weight approach and energy minimization model to solve the optimization problem,during which gives the corresponding data items different reward values to reduce apparent defective stripes,rested on the change of the initial disparity on the direction of column derived with dynamic programming algorithm with row,and finally acquires the final solution using dynamic programming algorithm with column.Experimental results show that the proposed algorithm can reduce the overall matching error rates and apparent defective stripes and get a better stereo matching effect.

关 键 词：立体匹配 行列动态规划 改进的自适应加权代价函数 奖励策略 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]