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.
出处
《电光与控制》
北大核心
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)
关键词
图像描述
图像重建
不变矩
重建误差
image description
image reconstruction
moment invariants
reconstruction error