基于指数矩的图像描述  

Image description based on exponent moments

作　　者：吴娱[1] 赵嘉济[1] 平子良[1] 杜昊翔[1] 

机构地区：[1]北京邮电大学世纪学院,北京102613

出　　处：《现代电子技术》2013年第14期112-115,共4页Modern Electronics Technique

基　　金：国家自然科学基金:窗口经验模式分解在图像处理中的应用(61065004)

摘　　要：论证一种基于T指数构建的圆谐傅里叶矩——指数矩(EFMs),分析其定义原理及其与基于三角函数构建的圆谐傅里叶矩的关系,验证指数矩作为一种正交不变矩所具有的多畸变不变性质。通过在Matlab软件平台上进行的仿真实验,证明了指数矩的旋转、缩放不变性,得出了指数矩作为一种高度浓缩的图像特征,无信息冗余,抽样性能好,抗噪声能力强,与其他矩相比更适用于多畸变不变图像描述和识别的结论。The circular harmonic-Fourier moments established on the basis of exponent, i.e. exponent moments （EMs） are demonstrated in this paper. The analysis of its principle and its relation with the CFMs verifies the multi-distorted invariance that EMs as the orthogonal invariant moments possess. Through a series of simulation experiments on Matlab platform, the rota- tion invariance and position invariance of EMs were proved. A conclusion that the EMs are more suitable for image description and recognition than other moments since it is a highly-concentrated image characteristic which has good sampling performance and strong anti-noise ability, but has no information redundancy.

关 键 词：数字图像 指数矩 图像归一化 不变量 

分 类 号：TN964-34[电子电信—信号与信息处理]