基于改进双边滤波的Kinect深度图像空洞修复算法研究

New Algorithm of Depth Hole Filling Based on Intensive Bilateral Filter

深度图像的灰度值表征的是物体与摄像头的距离,深度图像在三维场景重建、虚拟视点合成、3D视频技术中起到关键的作用。深度图像有多种获取方法,其中Microsoft公司的Kinect是一种廉价的深度图像获取设备,但产生的深度图像在物体的边缘和遮挡区域往往会产生空洞,不利于后续的研究应用。提出一种基于改进双边滤波法的深度图像修复算法,首先利用直方图确定阈值将深度图像二值化生成空洞Mask掩膜,然后利用改进的双边滤波法对空洞区域进行深度填充,最后利用中值滤波法进行图像平滑,实验结果表明该方法对于原始深度图像的空洞取得了较好的修复效果。

This paper presents a new method to fill the depth hole,first,a threshold is determined to binary the depth map by using a histogram,through this step the mask of the hole is gotten,then the depth hole is filled with the theory of intensive bilateral filter,and the depth map is filtered with a median filter.Experimental results show that the proposed algorithm can efficiently repair errors in the original depth map such as holes caused by occlusion.