摘要
通过计算图像二维直方图的非广延Tsallis熵,可以在去噪声的同时获取图像内部长程关联强度的信息。针对仅存在局域长程关联的红外线图片和无损检测图片,提出一种新的二维Tsallis熵判据,该方法把目标和背景当成相互独立的部分,能够最大限度保留二者各自信息的完整性。实验结果表明:该方法对非广延参数q的响应敏感度明显优于全局性长程关联假设下的方法,并且可对图像内部的局域长程关联进行定量描述。
By calculating nonextensive Tsallis entropy of two dimensional histogram of image,one can obtain intensity information of long-range correlation within image as well as denoising.A novel two dimensional Tsallis entropy criterion is proposed,aiming at those images presenting local long-range correlations,such as infrared ones and nondestructive test ones,in which object and background can be considered as two independent parts and information integrity of them can be maximized reserved by the proposed method.Experimental results show that the proposed method is more sensitive to nonextensive parameter q than that considering global long-range correlation,therefore,local long-range correlation hypothesis can be described quantitatively.
出处
《传感器与微系统》
CSCD
北大核心
2014年第7期150-153,共4页
Transducer and Microsystem Technologies
基金
国家自然科学基金资助项目(11005041)
福建省高校杰出青年科研人才培育计划资助项目(JA12001)
华侨大学中青年教师科研提升计划资助项目(ZQN-PY114)
关键词
图像分割
TSALLIS熵
长程关联
image segmentation
Tsallis entropy
long-range correlation
