【作者】 王国栋； 潘振宽； 赵希梅； 刘存良； 王钰；

【Author】 Wang Guo-dong~(1)),Pan Zhen-kuan~(2)),Zhao Xi-mei~(3)),Liu Cun-liang~(4)),Yu Wang~(5)) 1) 2) 3) 4)(College of Information Engineering,Qingdao University,Qingdao,266071) 5)(College of Mechanical & Electronic Engineering,Qingdao University,Qingdao,266071)

【机构】 青岛大学信息工程学院； 青岛大学机电工程学院；

【摘要】 针对传统的变分水平集方法在进行图像纹理分割时不能达到预期效果的问题,提出了一种基于图像分解与变分水平集相结合的分割方法。新提出的模型不需要引入如小波变换或者Gabor变换等辅助技术,而仅利用图像中的纹理信息和结构成份信息就能够达到纹理分割的效果。本文提出的分割模型通过凸松弛技术达到全局最优解,另外为了实现的方便,在求解时引入Split-Bregman方法,最后通过多组实验,验证了本文提出方法的有效性。更多还原

【Abstract】 In this paper we propose a new texture segmentation model based on image decomposition and active contour model.Unlike other active contour models for texture segmentation,our model doesn’t need any statistic model or other filters such as Gabor and Wavelet;we only use the structural and oscillating components’ information which can be estimated via variational image decomposition model.Our proposed model is global convex through global relaxation. Finally,the fast Split-Bregman algorithm of the proposed coupling model is designed for the segmentation implementation.The performance of our method is demonstrated through numerical experiments on some synthetic and real textured image.更多还原