改进Sobel算子在静态手势分割中的应用

摘要

为了解决传统的Sobel算法存在的斜方向不敏感问题, 本文提出了一种改进的Sobel算法并应用在静态手势分割中.该方法首先对静态图像双边滤波, 去噪且最大程度保留边缘;然后采用椭圆肤色模型筛选出肤色域, 根据肤色域特征分离手部;最后, 增加Sobel算法45°和135°两个方向模板, 重新分配权重, 使用改进算法提取手部轮廓.实验证明其在简单静态手势分割中具有较高精确度.%In order to solve the problem of the traditional Sobel algorithm being insenstive for the diagonal direction, an improved Sobel algorithm is proposed and applied to static gesture segmentation.The method firstly filters the static image, denoises and retains the edge to the greatest extent.Then the elliptical skin color model is used to filter out the skin color domain, and the hand is separated according to the skin color domain feature.Finally, the 45° and 135° direction templates of Sobel algorithm are added.Weights are reassigned and an improved algorithm is used to extract the hand contours.Experiments show that it has higher accuracy in the simple static gesture segmentation.