一种新的图像分割算法  

Novel image segmentation algorithm

作　　者：徐建军[1] 高山[1] 毕笃彦[1] 陈游[1] 

机构地区：[1]空军工程大学工程学院,陕西西安710038

出　　处：《西安电子科技大学学报》2011年第1期8-15,共8页Journal of Xidian University

基　　金：重点实验室基金资助项目(9140c610301080c6106)

摘　　要：针对一类普遍存在的图像,采用具有生物学背景的交叉视觉皮质模型进行图像分割.将交叉视觉皮质模型所具有的符合人眼对亮度响应非线性要求的指数衰减的阈值机制,改进为适合图像分割处理的线性衰减的阈值机制,提出了线性阈值-交叉视觉皮质模型.同时采用改进的二维Tsallis交叉熵作为分割准则,可自动地确定交叉视觉皮质模型神经元的分割阈值以及循环迭代次数.实验表明,这种分割算法优于经典的OSTU算法和K-m eans算法,同时基于改进的二维Tsallis交叉熵准则优于基于二维最大Shannon熵准则、传统二维Tsallis交叉熵准则和一维最小Tsallis交叉熵准则.The Intersecting Cortical Model （ICM） is adopted for image segmentation. Because the nonlinear exponential attenuation threshold mechanism is suitable for brightness respondenee but not suitable for image segmentation, we change it to the linear threshold mechanism which leads to the LT-ICM. Meanwhile, an improved 2-D Tsallis cross-entropy segmentation rule is proposed for determining the segmentation threshold and iteration time automatically. Experiments show that the proposed algorithm is better than OSTU and K-means algorithms for the texture image segmentation. Besides, our rule is more effective than the 2-D max-Shannonentropy rule, traditional 2-D Tsallis cross-entropy rule, and min-cross-entropy rule.

关 键 词：图像分割 交叉视觉皮质模型 二维Tsallis交叉熵 线性阈值 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]