Body Segmentation Based On Kinect Depth Camera

Body segmentation is to segment the human body from images which include human body.It is an important research direction in computer vision,and can be applied to many tasks,such as 3D body reconstruction,video editing,pose estimation problem and so on.However,body segmentation is challenging when background regions have similar colors to hairs,skin and/or clothes of the persons in the scene.To resolve efficient human body segmentation from complicated background,a body segmentation method from RGB-D and skeleton information based on the graph-based optimization framework is proposed in this paper.The algorithm considers mutual promotion of color,depth and skeleton information,and can get more accurate segmentation results.Firstly,this paper adopts an edge-guided filter to recover low quality depth map for obtaining high quality depth map.The original depth map captured by Kinect contains lots of noise and lack of depth value,which will influence the accuracy of image data.This paper adopts an edge-guided filter to handle low quality depth map for recovering accurate depth value.Then the RGB-D data is clustered into superpixels via a clustering algorithm.Finally,we propose a graph model,in which the superpixels are considered nodes and the associated skeleton is incorporated to enhance the capability of the graph in distinguishing body regions with similar color to the background.The data term and smoothness term of energy function is designed,and global optimal merge result will be gotten by minimizing the energy function.To evaluate the effectiveness of the proposed algorithm,several experiments on the real scenarios are compared.Experimental results show that the proposed method achieves more accurate body segmentation performance in both subjective visual and objective index comparisons.