高分辨率图像的指数矩描述  

High-Resolution Image Based on Exponent Moments

作　　者：孙景峰[1] 刘慧英[1] 纪超[1] 郭慧娟[1] 

机构地区：[1]西北工业大学自动化学院,西安710072

出　　处：《应用科学学报》2016年第2期127-134,共8页Journal of Applied Sciences

基　　金：国家自然科学基金(No.61371123;No.61301165)资助

摘　　要：在指数矩定义与计算方法的基础上,提出应用指数矩矩阵描述高分辨率图像的方法.结合图像分割理论并延伸处理各子图,利用快速傅里叶变换计算高分辨率图像指数矩矩阵,然后逆向重构子图并重组原图.对重组图像进行基于Itti视觉注意模型的客观感兴趣区域信噪比分析.仿真结果表明,延伸处理可减弱由分割所带来的重组图像方块效应,也可提高重构图像的信噪比,验证了由图像指数矩矩阵描述高分辨率图像的可行性.Based on the definition and calculation method of exponential moments, this paper describes a high-resolution image using the exponential moment matrix. Combining image segmentation and further treatment of sub-images, we use FFT to calculate the exponential moment matrix of a high-resolution image. Each sub-image is then reconstructed with the obtained exponential moment matrix. The original image is recomposed using the reconstructed sub-images. Meanwhile, based on Itti＇s visual attention model, SNR of the objectively interested area in the reconstructed image is analyzed. Simulation results show that the reconstructed image block effect caused by segmentation can be reduced, and SNR of the reconstructed image is improved. It is shown that describing a high-resolution image with image moment invariant matrix is feasible.

关 键 词：指数矩 图像归一化 图像质量 图像重构 

分 类 号：TP311[自动化与计算机技术—计算机软件与理论]