基于GVF Snake模型的舌体轮廓实时自动提取

Real-Time Automatic Extraction of Tongue Contours Based on GVF Snake Model

正确快速分割舌体轮廓是中医舌诊实时系统的重要前提。针对Snake算法中初始轮廓设置复杂,以及Snake模型因力场迭代次数过多而造成运行时间过长的问题,提出了一种新的舌体轮廓提取方法。该方法首先利用舌体的RGB值的特殊性,提取出增强的舌体区域的大致轮廓,然后计算出增强的舌体区域的梯度矢量流力场,最后根据该GVF力场的分布规律,自动寻找并获取到非常贴近初始轮廓的Snake轮廓线。实验表明,使用该方法,舌体轮廓分割能在2秒钟内完成,满足了中医舌诊系统的实时性。

Fast segmentation of tongue contours is an important prerequisite for traditional Chinese medicine tongue real-time diagnostic systems.Aiming at the complexity of setting initial contour,and long-running which is caused by excessive force field iterations in the snake algorithm,a new method of the tongue contour segmentation is proposed.This method starts from the specificity of the tongue＇s RGB value,and extracts out the broad contours of the enhanced tongue＇s regional,then,calculates the gradient vector flow force field.Finally,according to the distribution of the GVF force field,finds and obtains the snake contours automatically which is very close to the initial contour.Experiment shows that,in this method,the tonge contour segmentation can be completed in 2 seconds,and satisfies the real-time of the tongue diagnosis system.