基于肤色质心与边缘自生长的手势分割算法

针对现有手势分割方法难以在类肤色背景下从图像中高效完整地分割出静态手势的问题,提出一种基于肤色质心与边缘自生长的手势分割算法。利用肤色模型得到手势区域的质心,质心可降低后续边缘检测算法的计算量;利用改进的边缘检测算法得到手势边缘,同时提出一种边缘自生长算法,能有效补全局部断裂边缘,增强后续分割效果;将肤色信息与边缘信息进行差分运算分离类肤色背景,再用连通域及形态学处理去除以得到最终手势图像。实验结果证明,该算法较传统肤色模型及同类算法,能更加快速准确地在类肤色背景下分割出手势图像。

GESTURE SEGMENTATION ALGORITHM BASED ON COMPLEXION CENTROID AND EDGE SELF-GROWTH

Aimed at the problem that the existing gesture segmentation methods are difficult to efficiently and completely segment static gestures from images in skin-like backgrounds,a gesture segmentation algorithm based on skin color centroid and edge self-growth is proposed.The complexion model was used to obtain the centroid of the gesture areawhich could reduce the calculation amount of the subsequent edge detection algorithm.The improved edge detection algorithm was used to obtain the gesture edge.At the same time,an edge self-growth algorithm was proposed,which could effectively complement part of the fracture edge and enhance the subsequent segmentation effect.The skin color information and the edge information were subjected to a difference operation to separate the skin color-like background,and then connected domain and morphological processing were used to obtain the final gesture image.The experimental results prove that the proposed algorithm can segment the gesture image more quickly and accurately than the traditional skin color model and similar algorithms under the skin-like background.