基于优化水平集的细胞图像阈值分割算法

Threshold Segmentation Algorithm of Cell Image Based on Optimized Level Set

为实现对粘连细胞图像的分割,将Bayes分类器和KNN分类器引入到水平集外部速度函数的设计中,两种分类器轮流作用,无需设定阈值便能产生水平集驱动力.算法将Shi模型的双链表和C-V模型的全局分割相结合,以加快曲线演化.将目标与背景的类内平均距离引入到OTSU阈值法的阈值选择函数中,对OTSU法进行了改进.试验结果表明,相较于水平集法和阈值法,该算法对复杂粘连细胞的分割效果更好,在细胞图像分割中具备一定的有效性和可行性.

Due to the deficiency of fast level set edge detection method that requires manual setting of threshold value to obtain the driving force required by curve evolution,the method of pattern classification is introduced into the design of external velocity function.The bayesian classifier and the minimal neighbor classifier work alternately to generate the external driving force needed for curve evolution.The algorithm combines Shi model′s double-linked list with c-v model′s global segmentation to promote the curve′s rapid evolution.The OTSU method is improved by the average intra-class distance between the target and the background is introduced into the threshold selection function of OTSU.A large number of experimental results show that this algorithm has better segmentation accuracy and shorter running time.Compared with several classical level set segmentation algorithms and threshold method,the proosed algoritnm has a certain effectiveness and feasibisity in complex cell image segmentation.