基于图像融合的深度图像修复算法  

Depth image restoration algorithm based on image fusion

作　　者：刘汉伟 曹雏清 LIU Hanwei;CAO Chuqing(School of Mechanical Engineering,Nanjing University of Science&Technology,Nanjing 210094,China;HIT Wuhu Robot Technology Research Institute Co.,Ltd.,Wuhu 241007,China)

机构地区：[1]南京理工大学机械工程学院,江苏南京210094 [2]芜湖哈特机器人产业技术研究院有限公司,安徽芜湖241007

出　　处：《现代电子技术》2020年第2期182-186,共5页Modern Electronics Technique

基　　金：国家自然科学基金(61503186)

摘　　要：针对目前大多数深度相机采集到的深度图像中含有大量噪点以及大面积的空洞问题,提出一种基于图像融合的深度图像修复算法。采用改进分水岭算法提取彩色图像中的边缘信息,基于KD树近邻算法依据深度图像的梯度信息提取分类信息,将彩色图像的边缘信息与深度图像像素点的分类信息相结合,得到精确地图像分类结果,再对融合后的每一类进行最小二乘法算法拟合空洞,修复深度图像中出现的大面积空洞问题。实验结果表明,该方法在对物体边缘处小面积空洞进行较为准确地修复的同时,能够对深度图像中存在的大面积空洞问题进行有效修复。A depth image restoration algorithm based on image fusion is proposed to solve the problem that the depth images collected by most depth cameras contain a large number of noisy points and a large area of voids.The edge information of the color image is extracted by means of the improved watershed algorithm,and the classification information of the depth image is extracted accordingto the gradient information of it and based on the KD tree neighbor algorithm.The accurate image classification results are obtained in combination with the edge information of the color image and the classification information of the depth image pixels.For each class after fusion,the least square algorithm is used to fit the voids to repair the large⁃area voids in depth images.The experimental results show that the method can not only accurately repair small⁃area voids at the edge of the object,but also effectively repair large⁃area voids in the depth images.

关 键 词：深度图像修复 图像融合 提取边缘信息 图像分类 修复空洞 对比验证 

分 类 号：TN911.73-34[电子电信—通信与信息系统]