摘要
在经典的Chan-Vese模型中结合显著性分析,提出了一种有效的目标分割方法.即首先利用频谱残差方法提取图像的显著性区域,针对阈值分割方法的缺点使用改进的自适应阈值分割方法获取目标的大致轮廓,并以此轮廓作为Chan-Vese模型中初始曲线.该方法使得活动轮廓可以从靠近目标物体的位置进行演化,去除复杂背景的干扰.这样就解决了背景复杂时无法得到较为准确的边缘的问题;同时,也减少了CV模型的迭代次数.实验结果表明无论是背景复杂的灰度图像还是医学彩色图像,该算法的分割精度和运行效率都优于CV模型.
An effective object segmentation method is proposed which combines Chan-Vese model with saliency map.It firstly uses spectral residual method to extract the saliency region,then gets the general contours of the object using the improved adaptive threshold segmentation to aim at the shortcoming of the threshold segmentation.It makes this contours as the initial curves of Chan-Vese model.This method ensures the active contours evolve close to the object and removes the obstruction from the complex background.It solves the problem that accurate edge can＇t be obtained while the background of the image is complex;at the same time,it greatly reduces the number of iterations of the CV model.The experimental results show that the proposed algorithm is better than the CV method on precision and efficiency both for the complex gray image and medical color image.
出处
《微电子学与计算机》
CSCD
北大核心
2011年第10期21-23,27,共4页
Microelectronics & Computer
基金
江苏省高校自然科学基金计划项目(08KJB520001)
江苏省"青蓝工程"资助
淮安市科技项目(HAG2010030)
关键词
显著图
水平集
频谱残差
图像分割
saliency map
level set
spectral residual
image segmentation
