Abstract: Aiming at the conditions that the common edge detection algorithm is more sensitive to noise, obtained edge is not fine, and prone to false edge or edges overlapping, combined with traditional Chinese medicine tongue image feature, based on wavelet transform edge detection algorithm, improved wavelet transform edge detection algorithm is proposed. The algorithm calculates adjacent difference bit-by-bit based on the images of each row, column edge information, and detects wavelet extremum between potential difference. The experimental result proves that this method can effectively solve the contradiction between the traditional edge detection algorithm for removing noise and obtaining the fine edge, make the edges overlap phenomenon greatly reduced, thereby obtaining the ideal result of edge detection, laying a good foundation for the tongue area segmentation.

Key words: wavelet transform, traditional Chinese medicine, tongue image, edge detection

摘要： 针对目前常见的边缘检测算法对噪声较为敏感，获取的边缘不够精细，且容易出现伪边缘或边缘重叠等情况，结合中医舌象的特点，在小波变换边缘检测算法的基础上提出了改进的小波变换边缘检测算法。该算法通过对图像每一行、列的边缘信息逐位算出相邻位差，并在位差的差值变化大小之间检测小波极值。实验结果证明，该方法能有效解决传统边缘检测算法对去除噪声和获取精细边缘之间的矛盾，使边缘重叠现象大为减少，从而获得了比较理想的边缘检测效果，为以后整个舌体区域的分割提取打下了良好的基础。

关键词: 小波变换, 中医, 舌象, 边缘检测