基于像素滤波和中值滤波的深度图像修复方法  

Depth image inpainting method based on pixel filtering and median filtering

作　　者：刘继忠[1] 吴文虎 程承 王光辉[1,2] 曾成 LIU Ji-zhong;WU Wen-hu;CHENG Cheng;WANG Guang-hui;ZENG Chen(School of Mechanical ＆ Electrical Engineering, Nanchang University, Nanchang 330031,China;Department of ECCS, University of Kansas, Lawrence 66045, USA)

机构地区：[1]南昌大学机电工程学院,江西南昌330031 [2]堪萨斯大学电子工程与计算机科学系,堪萨斯劳伦斯66045

出　　处：《光电子．激光》2018年第5期539-544,共6页Journal of Optoelectronics·Laser

基　　金：国家自然科学基金(61273282);江西省科技计划(20161BBE50090);江西省自然科学基金(2017BAB201013)资助项目

摘　　要：针对Kinect相机采集的深度数据一般有噪声和黑洞现象,直接应用于场景三维重建系统中效果差的问题,提出了像素滤波器和中值滤波器相结合的Kinect深度图像修复方法。首先对一帧图片上的像素进行搜索,找到像素值为0的点,以该点为中心,利用其邻域内的像素,定义一个两层的滤波器,根据像素滤波器的原理对其进行修复,填充深度图片的黑洞;然后采用中值滤波器,在平滑深度图像的同时保留边缘信息,去除孤立噪声点。实验结果表明,本文方法去噪的同时也能对黑洞进行修补,与原始深度图像相比,空洞明显减少,深度图像质量大大提高。Generally,it has a poor effect if the depth data obtained by Kinect camera are directly used in 3D-reconstruction system due to the noise and black holes.To solve this problem,a depth image restoration method based on pixel filtering and median filtering is proposed.Firstly,search for each pixel in the picture and find the 0 value pixels.Secondly,use the pixel filter to repair it and fill the black holes in the picture.Finally,bilateral filters are used to remove noise and preserve edge information.The experimental results show that the proposed method can both remove the noise and repair the black holes at the same time.Compared with the original depth image,the black holes in the picture are obviously reduced,and the image quality is greatly improved.

关 键 词：KINECT 三维重建 深度图像修复 像素滤波器 中值滤波器 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]