融合对数极坐标和改进窄带法的图像分割方法  

Image Segmentation Method based on Combination of Log-polar Transform and Improved Narrow Band

作　　者：杨卫莉[1] 郭雷[1] 赵天云[1] 张冬梅[1] 

机构地区：[1]西北工业大学自动化学院,陕西西安710072

出　　处：《火力与指挥控制》2008年第2期8-12,共5页Fire Control & Command Control

基　　金：国家自然科学基金资助项目(60675015)

摘　　要：基于曲线演化的水平集算法近来已被广泛应用于图像分割中,针对其计算速度慢的问题,提出一种新的基于改进窄带法的图像分割方法INBM(Improved Narrow Band Method)。INBM首先将均匀采样的图像映射到对数极坐标系中,由视网膜空间分辨率机制可知,注视点都在图像兴趣区,由此形成初始轮廓,然后用改进的窄带水平集(Level Set)方法演化曲线得到最终分割结果。改进窄带法是通过降低窄带区域内的水平集函数求解个数,来减少计算时间。实验结果表明,该方法大大提高了图像分割的速度。Level set method based on curve evolution has been widely used in image segmentation. This paper presents a novel image segmentation method based on improved narrow band method （INBM） to solve the problem of low speed. The image is mapped from Cartesian to log-polar coordinate. From space variant theory of human vision system, that focus is on the interesting region, the initial contour of the region is formed. Using the improved narrow band level set method, curve is evolved to obtain result of image segmentation. The improved narrow band method adopts decreasing level set function solution to calculate number of narrow band region to reduce computing time. The experimental results show that the new method can speed up the image segmentation greatly.

关 键 词：图像分割 对数极坐标 水平集 窄带法 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]