面向多遮挡场景的光场深度估计  

Estimation of light field depth for multi-occlusion scenes

作　　者：吴英臣 张倩[1] 周超 杜昀璋 王斌[1] 严涛 WU Yingchen;ZHANG Qian;ZHOU Chao;DU Yunzhang;WANG Bin;YAN Tao(College of Information, Mechanical and Electrical Engineering, Shanghai Normal University, Shanghai 200234, China;School of Information Engineering, Putian University, Putian, Fujian 351100, China)

机构地区：[1]上海师范大学信息与机电工程学院,上海200234 [2]莆田学院信息工程学院,福建莆田351100

出　　处：《中国科技论文》2021年第7期715-722,共8页China Sciencepaper

基　　金：国家自然科学基金资助项目(61741111)。

摘　　要：为解决复杂场景中光场图像的遮挡问题,首先提出了一种改进的边缘检测方法,将Canny边缘检测与光流边缘检测相结合;其次提出了一种改进的层次聚类方法,添加孤立点检测,构建了基于类间平均聚合程度的相似性度量,通过优化类间平均聚合程度和类内集中程度,以确定最终聚类结果;最后计算每个区域的匹配代价,以获得深度图。实验结果表明,所提出方法在多遮挡场景的光场深度估计中能生成视觉质量高、保真度强的深度图像,优于当前几种流行的算法。In order to solve the occlusion problem of light field image in complex scenes,a novel method to solve the occlusion problem of the light field image in multi-occlusion scenes was proposed.Firstly,an improved edge detection method was proposed by combining Canny with optical flow edge detection.Secondly,an improved hierarchical clustering method was proposed with adding outlier detection and constructing a similarity measure based on the average degree of aggregation between clusters.The final clustering result was determined by optimizing the average degree of attraction between clusters and the degree of intra-cluster concentration.Finally,the depth maps were obtained by calculating the matching cost for each region.Experimental results show that the proposed method can generate high-quality and high-fidelity depth images in the light field depth estimation of multi-occlusion scenes,which is superior to several popular algorithms.

关 键 词：光场 深度估计 重聚焦 遮挡边缘检测 遮挡区域分类 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]