基于改进交叉视觉皮质模型的图像分割新方法  

New image segmentation method based on improved intersecting cortical model

作　　者：周理[1] 高山[1] 毕笃彦[1] 何林远[1] 

机构地区：[1]空军工程大学航空航天工程学院,陕西西安710038

出　　处：《计算机工程与科学》2013年第7期124-130,共7页Computer Engineering & Science

基　　金：航空科学基金资助项目(20101996009);国防科技重点实验室基金资助项目(9140C610301080C6106);国家自然科学基金资助项目(61175029)

摘　　要：从人类视觉特性出发,将具有"中心抑制,周边加强"性能的人眼视觉侧抑制现象与交叉视觉皮质模型相结合,形成了一种与人类视觉保持一致的适合图像分割的侧抑制-交叉视觉皮质模型。在此基础上,利用灰度级-邻域平均灰度级二维直方图,将一维卡方散度推广至二维,构造出了基于最小二维卡方散度的分割准则,进而提出一种实时性高的图像分割新算法。实验仿真表明,这种分割算法效果优于原始ICM算法和经典的Otsu算法,同时该准则计算速度远高于基于二维最小交叉熵准则的计算速度。In the view of human visual feature, the Lateral Inhibition of human vision that has fea- tures of inhibiting center and enhancing surroundings is combined with Intersecting Cortical Model, for the sake of forming the Lateral Inhibition-Intersecting Cortical Model which brings into correspondence with human visual. Meanwhile, through the gray level-average gray level histogram, 1-D Chi-square di- vergence is extended to 2-D Chi-square divergence, then the segmentation criterion based on the rule of 2-D Chi-square divergence is structured. Further, a new real-time segmentation algorithm is proposed. At last, simulations validate that this algorithm is not only more simple than 2-D cross-entropy one in computing complexity, but also superior to the classical original ICM and Otsu in the performance of outcomes.

关 键 词：交叉视觉皮质模型 灰度级-邻域平均灰度级直方图 二维卡方散度 侧抑制模型 抑制系数 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]