基于最小Tsallis交叉熵的阈值图像分割方法  

Method for thresholding image segmentation based on minimum Tsallis-cross entropy

作　　者：唐英干[1] 邸秋艳[1] 关新平[1] 刘福才[1] 

机构地区：[1]燕山大学工业计算机控制工程河北省重点实验室,秦皇岛066004

出　　处：《仪器仪表学报》2008年第9期1868-1872,共5页Chinese Journal of Scientific Instrument

基　　金：国家杰出青年基金(60525303);燕山大学博士基金(B243)资助项目

摘　　要：利用Tsallis熵的非广延性(nonextensivity),本文提出了一种最小Tsallis交叉熵阈值分割方法。该方法通过引入一个参数q,在阈值的选择过程中不仅考虑了目标和背景之间的信息量差异,而且考虑了目标和背景之间的相互关系,克服了传统最小交叉熵忽略目标和背景之间的相互关系所导致的阈值选择不恰当的缺点。实验结果表明,通过选择合适的参数q,所提出方法的分割效果明显优于传统的最小交叉熵阈值法。Utilizing the nonextensivity of Tsallis entropy, a thresholding segmentation method based on minimum Tsallis-cross entropy is proposed in this paper. A parameter q is introduced in this method, which not only considers information discrepancy between the object and the background in the process of choosing image threshold, but also considers the relationship between the two classes. The proposed approach conquers the flaw of choosing an improper threshold of the conventional minimum cross-entropy method. Experimental results show that the proposed method performs better than the conventional minimum cross-entropy threshold selection by choosing a proper value of parameter q.

关 键 词：图像分割 阈值 最小交叉熵 TSALLIS熵 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]