融合GVF Snake与肤色模型的手势轮廓提取方法  

Fusion Extraction Method of Gesture Contour Based on GVF Snake and Skin Color Model

作　　者：覃文军[1] 杨金柱[1] 宋相满[2] 赵大哲[1] 

机构地区：[1]东北大学医学影像计算教育部重点实验室,沈阳110819 [2]东北大学信息科学与工程学院,沈阳110819

出　　处：《小型微型计算机系统》2013年第6期1405-1408,共4页Journal of Chinese Computer Systems

基　　金：国家自然科学基金项目(61001047;61172002)资助;中央高校基本科研业务费项目(N110804005)资助

摘　　要：针对在基于视觉的手势识别系统中手势轮廓难以准确提取问题,本文提出一种融合GVF Snake和肤色模型的手势轮廓提取方法.首先把图像由RGB空间转换到YCb'Cr'空间,利用该空间上的椭圆肤色模型检测出手势区域并提取手势轮廓作为GVF Snake模型的初始轮廓曲线;然后根据图像分块思想把检测出的手势所在图像区域分割出来,并计算该图像分块的梯度值;最后在图像分块和初始轮廓曲线的基础上通过GVF Snake模型迭代搜素准确提取手势轮廓.实验结果表明,本文提出的手势轮廓提取方法无需人工参与,准确性上优于肤色模型、传统Snake模型,实时性上优于GVF Snake模型,满足手势识别系统中手势轮廓提取的实时性和准确性要求.检测准确、实时性高.For the hand gesture contour of the vision-based gesture recognition system is difficult to accurately extracted, this paper presents a fusion extraction method of hand gesture contour based on GVF snake and skin color model. Firstly, the image is converted from RGB color space to YCb＇Cr＇ color space, and the gesture contour is detected by elliptical skin color model in YCb＇Cr＇ color space as the initial profile curve for GVF Snake model. Then the image block of the gesture is separated and the gradient of the image block is computed. Finally, the accurate gesture contour is extracted by iterative search with GVF Snake model on the basis of the im- age block and the initial profile curve. Experimental results show that the proposed method of gesture contour extraction is without hu- man intervention, the detection accuracy is better than the skin color model and Snake model, and the real-time is higher than the GVF Snake model. The proposed method can meet the accuracy and real time request of hand gesture contour extraction in the hand gesture recognition system.

关 键 词：手势识别 手势轮廓 肤色模型 SNAKE模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]