一种快速红外图像分割方法_杜峰.pdfVIP

24 3 Vo.l 24, No. 3 2005 6 J. InfraredM illim. W aves June, 2005 :1001- 9014(2005)05 -0370 - 04 1 1 1 2 杜　峰 , 　施文康 , 　邓　勇 , 　朱振幅 (1. , 　200030; 2. 207 , 　100854) :为准确地实现目标识别, 提出了将二维最大熵图像 割方法应用于红外图像实行 割. 利用图像的二维直方 图, 二维最大熵 割方法不仅考虑了象素的灰度信息, 而且还充 利用了象素的空间领域信息, 能取得较为理想的 割结果. 然而该方法所需的巨大运算量限制了其实际应用. 运用 SO 算法代替穷尽搜索获得阈值向量, 求解速度 可提高 300 ～400倍, 提高了 割效率. 通过对实际的红外图像 割表明, 这种方法简单、有效. 　　:图像 割;微粒群优化;熵;目标识别 :T 391. 4　　:A FAST INFRARED IMAGE SEGMENTATION METHOD 1 1 1 2 DU Feng, 　SH IWen-K ang , 　DENG Yong, 　ZHU Zhen-Fu (1. Inst of Au tom atic Detection Technology, Shanghai Jiao Tong University, Shanghai 200030, China; 2. Laboratory of Target and Environm ent Feature, CASIC, Beijing 100854, China) Abstrac:t In order to detect objectsaccurately, an mi age thresholding approach named wt o dmi ensions (2-D)m aximum en- tropy was proposed to do infrared image segmentation. By using the 2-D histogram of mi age, the 2-D m axmi um entropy m ethod not only considers the distribu tion of gray inform ation, but also takes advantage of the spatial neighbor inform ation. How ever, its great com putationwasoften an obstacle in application. The th reshold vectorwas obtained by using a new opt-i m ization algorithm, namely, the particle sw arm optim ization algorithm ( SO). The new w ay w asp roposed to realize the 2- D m aximum entropym ethod instead of exhaustive search m ethod. And it is 300 ～400 tmi es faster than the traditionalmeth- od. Through the exam ple of segm enting the infrared im age, the p roposed m ethod has been proved to be a fast method of