模拟退火优化fcm 聚类高光谱图像压缩研究 - 北京邮电大学学报.pdfVIP

2015年10月 北 京 邮 电 大 学 学 报 Oct.2015 第38卷 第5期 Journal of Beijing University of Posts and Telecommunications Vol.38 No.5 摇 摇 文章编号:1007鄄5321(2015)05鄄0058鄄04 DOI:10.13190/ j.jbupt.2015.05.010 模拟退火优化FCM聚类高光谱图像压缩研究 赵学军,摇 王晓娟,摇 于凯敏,摇 乔摇 旭 (中国矿业大学(北京) 机电与信息工程学院,北京 100083) 摘要:基于矢量量化的高光谱图像有损压缩算法可以获得较高的压缩比,但是其时间复杂度高,失真较大. 为此, 提出了模拟退火优化模糊C均值聚类(FCM)的高光谱图像有损压缩算法. 先对高光谱图像进行自适应波段合并 算法降维,利用肘部现象确定量化级数,结合模拟退火的全局寻优能力和模糊聚类的快速收敛能力,找到最优解后 恢复维度,最后去模糊优化编码方案. 通过这种方法,在提高高光谱图像压缩运算效率和减小解压后失真方面都有 了较大的优化,是基于矢量量化的高光谱图像压缩的可行方法. 关摇 键摇 词:模拟退火;模糊聚类;降维;肘部现象;矢量量化 中图分类号:TP751郾1摇 摇 摇 摇 文献标志码:A A New FCM Based on Simulated Annealing of Hyperspectral Image Compression ZHAO Xue鄄jun,摇 WANG Xiao鄄juan,摇 YU Kai鄄min,摇 QIAO Xu (School of Mechanical Electronic and Information Engineering,China University of Mining and Technology (Beijing), Beijing 100083,China) Abstract:Lossy compression of hyperspectral image based on vector quantization algorithms can achieve a high compression ratio,but it is of time complexity and great distortion. This article proposed a new fuzzy C鄄means clustering (FCM) algorithm based on simulated annealing. Firstly,the dimensions were reduced by using the algorithm of adaptive band combination dimensional reduction (ABC),then the number of clusters with the elbow was determined. FCM was combined with simulated annealing,and found optimal result quickly,then recovered dimensions. We got optimization coding by deblurring U. Through this approach,the efficiency has been improved and the distortions have been reduced greatly. Key words:simulated annealing;fuzzy C鄄means clustering ;dimension reduction;the elbow;