基于改进链码的新型手形图像分割方法

Improved Freeman chain code for hand shape segmentation

为精确定位手形图像中的分割拐点并避免检测出伪拐点,改进了基于链码的拐点检测方法,并应用于手形图像的指尖、指跟点定位。采用多阶差分方法,改进基于Freeman链码的拐点检测方法,以去除伪拐点和精确定位拐角点;对手形图像进行二值化和边缘提取后,利用改进的链码法检测手形轮廓点中具有局部曲率极值的拐点,作为手形分割的特征点。通过实验对比,该方法在手形分割和识别中达到了较好的效果,身份识别率达到95.95%。

An improved corner detection algorithm based on Freeman chain code was developed to improve the ability oI hand shape segmentation. The k-order difference method is adopted to improve corner detection algorithm to eliminate the spurious corners and locate true corners. After binarized and edge extraction, detected the local curvature extrema corner point of hand shape contour point as the feature points of hand shape segmentation. Hand shape segmentation experiments proved the robustness of the proposed method. Experimental results from the hand shape segmentation and recognition are very promising with the best recognition result of 95.95 %.