基于正交多项式不变矩的图像描述  

Image Description Using Orthogonal Polynomial Moment Invariants

作　　者：高生强[1] 杨珺[2] 熊广芝[1] 任华锋[1] 吉楠[1] 

机构地区：[1]空军工程大学理学院,西安710051 [2]西安理工大学自动化学院,西安710048

出　　处：《电光与控制》2009年第2期48-50,71,共4页Electronics Optics & Control

基　　金：国家自然科学基金资助项目(60573040)

摘　　要：通过对大图像、小图像、噪声图像的重建,比较了泽尼克矩、正交的傅里叶-梅林矩,畸变的雅可比-傅里叶矩的图像描述能力,最后得出:畸变的雅可比-傅里叶矩有着最强的图像描述能力。在实验中还发现:在噪声图像的重建中,随着重建阶数的提高,图像的重建误差并不是一直减少,而是和有噪声图像一样,是一个先降后升的过程,并对此现象作了解释:在离散空间中连续正交多项式矩并不是完全意义上的正交,是这种正交误差造成了此现象。Through reconstruction of large, small and noise-infected images, the image description capability of Zernike moment invariants, orthogonal Fourier-Mellin moment invaraints and Pseudo Jacobi moment invaraints were studied and compared with each other. The conclusion showed that the Pseudo Jacobi-Fourier moment has the best description result. It was also discovered that： in the reeonstuction of non-noisy images, the reconstruction error did not decrease continuously with the increasing of reconstruction order, instead, it decreased at first and then increased, like that of a noisy image. An explanation was given.

关 键 词：图像描述 图像重建 不变矩 重建误差 

分 类 号：V271.4[航空宇航科学与技术—飞行器设计] TP391.41[自动化与计算机技术—计算机应用技术]