摘要
基于在保证实时性的前提下提高匹配精确度的目的,本文采用了一种基于扩展双权重聚合的实时立体匹配方法。扩展双权重聚合是基于以下2个新的想法。第一,扩展双权重聚合越过颜色边界连接相似区域,对于一个给定的像素给予一个很大的支持区域。第二,它不包括支持区域中的预估异常值,从而制作一个高质量的支持区域。通过实验将扩展双权重聚合的实时立体匹配方法和其他实时立体匹配方法的精确性相比较。实验结果表明,所提出的立体匹配方法显示了一个平均5.10%的坏像素率,是这些实时立体匹配方法中精确度最高的。
In order to ensure the accuracy of real-time to improve the matching accuracy, this paper puts forward a real-time stereo matching method based on extended binary weighted aggregation.The extended binary weighted aggregation is based on the following two new ideas.First, the extended binary weighted aggregation connects distant regions over color boundaries to make them one large support region for a given pixel.Second, it excludes outliers in the support region to make a high quality support region. The extended binary weighted aggregation and other real-time stereo matching methods are compared by the experiment. Experimental results show that, the proposed stereo matching method shows an average of 5.10% of the bad pixel rate, it is the highest accuracv of these all existing real-time stereo matching methods.
出处
《电子设计工程》
2017年第6期160-164,169,共6页
Electronic Design Engineering
关键词
立体匹配
双权重聚合
远程连接聚合
异常排除聚合
stereo matching
binary weighted aggregation
remotely connected aggregation
outlier-excluded aggregation
