基于Marr-Hildreth算子的改进人体轮廓提取方法  

An Improved Human Body Contour Extraction Method Based on Marr-Hildreth Operator

作　　者：吴鹏飞 郭晨 WU Pengfei;GUO Chen(Wuhan Digital Engineering Institute,Wuhan 430205)

机构地区：[1]武汉数字工程研究所,武汉430205

出　　处：《计算机与数字工程》2024年第10期3126-3130,共5页Computer & Digital Engineering

摘　　要：传统人体轮廓提取大多基于RGB图和边缘检测算子,但RGB图受环境影响较大,Marr-Hildreth和Canny等算子也会减弱人体边缘信息,无法有效区分人体与背景。针对以上不足,基于Kinect获取的深度图,提出一种改进的人体轮廓提取方法。首先从Kinect获取人体深度图,利用中值滤波和保边高斯滤波组成联合滤波器来替代传统Marr-Hildreth算子中的高斯滤波器,最后对滤波图像做拉普拉斯和零交叉点检测处理,得到人体边缘轮廓。实验结果表明,论文算法能有效的提取人体轮廓,且边缘点的保护效果更好。Traditional human contour extraction is mostly based on RGB images and edge detection operators,but RGB imag-es are greatly affected by the environment.Operators such as Marr-Hildreth and Canny will also weaken the edge information of the human body and cannot effectively distinguish the human body from the background.In view of the above deficiencies,an improved human contour extraction method is proposed based on the depth map obtained by Kinect.First,the human depth map is obtained from Kinect,median filter and edge-preserving Gaussian filter is used to form a joint filter to replace the Gaussian filter in the tradi-tional Marr-Hildreth operator,and finally Laplacian and zero-cross point detection processing are performed on the filtered image,the outline of the human body edge is gotten.The experimental results show that the algorithm in this paper can effectively extract the outline of the human body,and the protection effect of edge points is better.

关 键 词：深度图 Marr-Hildreth算法 联合滤波器 图像边缘信息 

分 类 号：TP301.6[自动化与计算机技术—计算机系统结构]