利用图像相位信息构造的仿射-卷积矩不变量  

Affine and Convolution Moment Invariants Using Phase Information of Images

作　　者：公明 曹伟国[1] 李华[1,2] 

机构地区：[1]中国科学院计算技术研究所智能信息处理重点实验室,北京100190 [2]中国科学院大学,北京100049

出　　处：《计算机辅助设计与图形学学报》2014年第6期851-860,共10页Journal of Computer-Aided Design & Computer Graphics

基　　金：国家"八六三"高技术研究发展计划(2007AA01Z317);国家自然科学基金(60873164;61379082)

摘　　要：不理想的成像设备或者成像条件经常会导致图像的退化现象,如相机失焦、相机与场景的相对运动等会使得图像模糊,不同拍摄视角会引起图像中几何形状的变化等.为了实现对退化图像的识别、检索等,利用图像的相位信息构造了一种仿射-卷积矩不变量对这些退化图像进行描述.首先利用高斯卷积变换、仿射变换对图像的模糊、几何变化建模;然后分析图像的相位谱图像在高斯卷积变换、仿射变换下的性质,提出了能同时对高斯卷积变换和仿射变换具有不变性质的不变量计算方法.通过对计算机合成变换的图像、真实拍摄的图像进行检索、识别实验,表明了文中提出的仿射-卷积矩不变量的有效性和适用性.The observed images are usually degraded by non-ideal imaging systems or imaging conditions,for example,images are likely to contain blur due to being out of focus or relative motion between the camera and the scene,geometric degradations are often present in images due to different viewpoints.To solve this problem,a kind of combined moment invariants to affine transformation and Gaussian convolution degradation is proposed.First,blur and geometric degradation by Gussian convolution and affine transformation are modeled.Then,the property of the Fourier spectrum after combined affine and Gaussian transforms are analysed,and a kind of combined affine and convolution invariants in explicit forms are proposed.Through retrieval and recognition experiments to images under supervised transformations by computer and those shot in real scenes,the effectiveness of these combined invariants in the degraded scene is validated.

关 键 词：矩不变量 图像模糊 几何退化 卷积变换 仿射变换 图像相位 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]