文献类型：期刊文献

中文题名：基于改进的种子点扩展自适应立体匹配算法

英文题名：ADAPTIVE STEREO MATCHING ALGORITHM BASED ON IMPROVED SEED POINTS EXPANSION

作者：李鹏格[1];戴本祁[1]

机构：[1]华东理工大学信息科学与工程学院,上海200237

年份：2016

卷号：33

期号：3

起止页码：217

中文期刊名：计算机应用与软件

外文期刊名：Computer Applications and Software

收录：CSTPCD;;CSCD:【CSCD_E2015_2016】；

语种：中文

中文关键词：交叉自适应窗口;自适应权重;孤立点;种子点

外文关键词：Cross-adaptive window; Adaptive weight; Isolated point; Seed point;

摘要：由于在图像中不同区域的纹理密集程度不同,因此使用固定窗口大小的算法无法兼顾纹理不同的区域,并且在视差不连续区域的匹配精度较低。针对该问题,提出一种自适应窗口和自适应权重相结合的算法,并且采用种子点扩展的方法。首先,通过交叉自适应窗口法,区分出连续点和孤立点,对于不同的分类点采用不同的处理方法。其次,针对每一个像素点,利用改进的自适应权重方法进行匹配。最后,提出一种新的种子点扩展的视差优化方法,对初始视差图进行精细化。实验结果表明,视差图中纹理密集区域和视差不连续区域的误匹配现象得到改善。该算法可以有效地处理图像中纹理分布不均的问题,提高了在视差不连续区域内匹配精度。
Since the image in different regions has different texture intensity,therefore the algorithm using a size-fixed window can＇t take into account the regions with different textures,and its matching precision is low in disparity discontinuities regions. To solve these problems,in this paper we propose a new algorithm which combines adaptive window and adaptive weight,and uses the means of seed point expansion.First,we distinguish the consecutive points and isolated points with cross-adaptive window method,and use different processing approaches for different classification points. Secondly,for each pixel point we use the improved adaptive weight method to match it. Finally,we present a new disparity optimisation approach of seed points expansion to refine the raw disparity map. Experimental results show that the phenomenon of mismatching the texture intensity region and disparity discontinuities region in disparity map has been improved. The algorithm can effectively process the problem of uneven texture distribution in image,and raise the matching precision in disparity discontinuities region.