针对近年提出的二维Arimoto 熵阈值分割方法只依赖图像灰度级出现的频数信息，而未考虑图像类内灰度均匀性这一问题，提出了基于灰度-梯度直方图的二维Arimoto 灰度熵阈值分割方法. 首先，在Arimoto 熵的基础上直接考虑图像类内灰度均匀性，构建出一维Arimoto 灰度熵阈值选取公式；结合灰度-梯度二维直方图目标与背景区域划分方式，推导出二维Arimoto 灰度熵阈值选取公式；通过阈值选取函数所涉及中间变量的递推计算公式来消除冗余计算；采用基于Tent映射的混沌序列对人工蜂群算法的局部搜索阶段进行改进，以改进后的蜂群优化算法来加快图像分割最佳阈值的搜索速度，大大减少了时间花费. 大量的典型图像对比实验结果表明，所提出的方法能够快速而准确地实现图像分割，且总体效果优于二维Shannon 熵、二维Tsallis 灰度熵和二维Arimoto熵阈值分割方法.
A recently proposed 2D Arimoto entropy thresholding method only depends on frequency information of gray scale in an image, without considering uniformity of within-class gray scales. To solve this problem, a 2D Arimoto gray entropy thresholding method based on gray scale-gradient histogram is proposed.Uniformity of within-class gray scale is considered based on Arimoto entropy and a formula for 1D Arimoto gray entropy threshold selection constructed. Using regional division of object and background in a gray scale-gradient 2D histogram, a formula for 2D Arimoto gray entropy threshold selection is derived. Recursion formulae of intermediate variables in the threshold selection criterion function are used to eliminate redundant computation. The local period of an artificial bee colony algorithm is improved using a chaotic sequence based on tent mapping. The improved bee colony optimization algorithm can accelerate search speed of the optimal threshold for image segmentation to significantly reduce execution time. Experimental results based on a large number of typical images show that the proposed method can segment image quickly and accurately, with the overall performance better than 2D Shannon entropy thresholding, Tsallis gray entropy thresholding, and Arimoto entropy thresholding.
[1] CHANG C I, DU Y, WANG J, et al. Survey and comparative analysis of entropy and relative entropy thresholding techniques[J]. Vision, Image and Signal Processing, 2006, 153(6): 837-850.
[2] LINDA Mahmoudi, ALI El Zaart. A survey of entropy image thresholding techniques[C]//2012 2nd International Conference on Advances in Computational Tools for Engineering Applications. Beirut, 2012. 204-209.
[3] BARDERA A, BOADA I, FEIXAS M, et al. Image segmentation using excess entropy[J]. Journal of Signal Processing Systems, 2009, 54(1-3): 205-214．
[4] KAPUR J N, SAHOO P K, WONG A K C. A new method for grey-level picture thresholding using the entropy of the histogram[J]. Computer Vision, Graphics and Image Processing, 1985, 29(3): 273-285.
[5] SAHOO P K, ARORA G A. Thresholding method baded on two-dimensional renyi's entropy[J]. Pattern Recognition, 2004, 37(6): 1149-1161.
[6] 黄金杰，郭鲁强，逯仁虎，等. 改进的二维Renyi熵图像阈值分割[J]. 计算机科学，2010, 37(10): 252-253.
HUANG Jinjie, GUO Luqiang, LU Renhu, et al. Image threshold segmentation based on improved two-dimensional Renyi entropy[J]. Computer Science, 2010, 37(10): 252-253. (in Chinese)
[7] TANG Yinggan, DI Qiuyan, GUAN Xingping. Fast recursive algorithm for two-dimensional Tsallis entropy thresholding method[J]. Journal of Systems Engineering and Electronics, 2009, 20(3): 619-624.
[8] ARIMOTO S. Information theoretical consideration on estimation problems[J]. Information and Control, 1971, 19(3): 181-194.
[9] ZHANG Hong. One-Dimensional Arimoto entropy threshold segmentation method based on parameters optimization[C]// International Conference on Applied Informatics and Communication. Xi'an, China. 2011. 573-581
[10] 卓问，曹治国，肖阳. 基于二维 Arimoto 熵的阈值分割方法[J]. 模式识别与人工智能，2009, 22(2): 208-213.
ZHUO Wen, CAO Zhiguo, XIAO Yang. Image thresholding based on two-dimensional Arimoto entropy[J]. Pattern Recognition and Artificial Intelligence, 2009, 22(2): 208-213. (in Chinese)
[11] LIU Yaoyong, LI Shuguang. Two-dimensional Arimoto entropy image thresholding based on ellipsoid region search strategy[C]// 2010 International Conference on Multimedia Technology (ICMT). Ningbo China, 2010. 1-4.
[12] 张弘，范九伦. 二维Arimoto熵直线型阈值分割法[J]. 光子学报，2013, (2): 234-240.
ZHANG Hong, FAN Jiulun. Two-dimensional Arimoto entropy linear-type threshold segmentation method[J]. Acta Photonica Sinica, 2013, (2): 234-240. (in Chinese)
[13] KARABOGA D. An idea based on honey bee swarm for numerical optimization[R]. Kayseri: Engineering Faculty Computer Engineering Department, Ereiyes University, 2005.
[14] 刘三阳，张平，朱明敏. 基于局部搜索的人工蜂群算法[J]. 控制与决策，2014, 29(1): 123-128.
LIU Sanyang, ZHANG Ping, ZHU Mingmin. Artificial bee colony algorithm based on local search[J]. Control and Decision, 2014, 29(1): 123-128. (in Chinese)
[15] 杜峰，施文康，邓勇. 一种快速红外图像分割方法[J]. 红外与毫米波学报，2005, 24(3): 370-373.
DU Feng, SHI Wenkang, DENG Yong. Fast infrared image segmentation method[J]. Journal Infrared Millimeter and Waves, 2005, 24(3): 370-373. (in Chinese)
[16] 吴一全，吴诗婳， 张晓杰. 利用混沌 PSO 或分解的 2 维 Tsallis 灰度熵阈值分割[J]. 中国图象图形学报，2012, 17(8): 902-910.
WU Yiquan, WU Shihua, ZHANG Xiaojie. Two-dimensional Tsallis gray entropy image thresholding using chaotic particle swarm optimization or decomposition[J]. Journal of Image and Graphics, 2012, 17(8): 902-910. (in Chinese)
[17] LAVENDA B. H, DUNNING Davies J. Qualms concerning Tsallis’s condition of pseudo-additivity as a definition of non-extensivity[EB/OL].http://arxiv.org/abs/cond-mat/0311477v1, Nov-20, 2003.
[18] 吴一全，张金矿. 改进的2维 Otsu 法及混沌粒子群递推的阈值分割[J]. 中国图象图形学报，2009, 14(9): 1843-1849.
WU Yiquan, ZHANG Jinkuang. Thresholding based on improved 2D Otsu method and chaotic particle swarm optimization[J]. Journal of Image and Graphics, 2009, 14(9): 1843-1849. (in Chinese)
