基于四梯度多特征和四权重滤波的立体匹配  

Stereo Matching Based on Quad Gradient Multi-Feature and Quad-Weight Filtering

作　　者：董振武 王春媛 余嘉昕 周斌 吴扬[3] DONG Zhenwu;WANG Chunyuan;YU Jiaxin;ZHOU Bin;WU Yang(School of Electronic and Electrical Engineering,Shanghai University of Engineering Science,Shanghai 201620,China;Research Institute of Shanghai HaiChao New Technology,Shanghai 200070,China;Shanghai Institute of Satellite Engineering,Shanghai 200240,China)

机构地区：[1]上海工程技术大学电子电气工程学院,上海201620 [2]上海海潮新技术研究所,上海200070 [3]上海卫星工程研究所,上海200240

出　　处：《计算机工程》2025年第4期271-280,共10页Computer Engineering

基　　金：国家自然科学基金(61801286);上海市科技创新重点项目(22DZ1100803)。

摘　　要：立体匹配的精度直接决定了后续三维场景信息恢复的准确度,提高视差图的精度一直是研究的热点问题。传统的立体匹配方法对图像的局部结构表达不够精准,对相似结构区域内的点或前景和背景交界处及含有错误代价点的处理效果不理想。针对以上问题,提出一种基于四梯度多特征代价和四权重滤波的立体匹配方法。由图像灰度和四方向梯度组成多特征空间,通过二次编码计算多特征Census变换代价,再结合多特征绝对误差(AD)代价形成多特征代价,提升局部结构的表达准确度;以空间邻近度、像素值相似度、区域相似度和代价相似度四权重构建的滤波核进行代价聚合,降低异常代价的聚合权重;以胜者为王(WTA)算法计算初始视差,以左右一致性检测进行视差初步修正,结合自适应窗口和视差阈值进行视差优化。在Middlebury V3立体平台上的实验结果表明,该方法在非遮挡区域和全部区域加权平均的bad4.0分别为14.7%和20.6%,效能显著优于现有的传统立体匹配算法。The accuracy of stereo matching directly determines the precision of subsequent 3D scene information recovery,and enhancing the accuracy of disparity maps has attracted considerable attention among researchers.Traditional stereo matching methods inadequately represent the local structure of images,particularly within regions of similar structures,at the junctions of the foreground and background,and in areas containing erroneous cost points.To address these issues,this paper proposes a stereo matching method based on a quadgradient,multi-feature cost,and quad-weight filtering.This method constructs a multi-feature space composed of image intensity and gradients in four directions.It employs quadratic encoding to calculate the cost of the multi-feature census transform of images and then combines it with the multi-feature Absolute Difference(AD)cost to enhance the accuracy of the local structural representation.A filter kernel constructed with four weights,namely,spatial proximity,pixel intensity similarity,regional similarity,and cost similarity,is used for cost aggregation to mitigate the aggregation weight of abnormal costs.The initial disparity is calculated using the Winner-Take-All(WTA)algorithm and is preliminarily corrected through a left-right consistency check,followed by disparity optimization using an adaptive window and a disparity threshold.Results of experiments on the Middlebury V3 stereo platform indicate that the algorithm significantly outperforms existing traditional stereo matching algorithms.It yields a weighted average bad4.O value(percentage of"bad"pixels having an error greater than 4.0 pixels)of 14.7%in non-occluded regions and 20.6%in all regions.

关 键 词：立体匹配 绝对误差(AD)-Census变换 四梯度多特征 双边滤波 胜者为王(WTA)算法 代价聚合 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]