基于不变矩相似度的图像质量评价方法  

Image Quality Assessment Based on Similarity of Invariant Moments

作　　者：潘国华[1] 

机构地区：[1]烟台市公安局刑事科学技术研究所,山东烟台264003

出　　处：《烟台职业学院学报》2011年第4期51-56,共6页Journal of Yantai Vocational College

基　　金：国家自然科学基金项目资助(NO:60541001);烟台市科技规划项目(NO:2010114)

摘　　要：针对图像质量评价问题,考虑算法对不同大小和各类变形图像的普适性,提出了一种基于Hu不变矩相似度的图像评价方法。该方法通过计算图像的7个Hu不变矩值,进而计算两图像不变矩距离,利用不变矩距离大小来度量两图像的相似程度,从而实现了对待测图像的质量评价.实验结果表明:对于尺寸相等的待测图像与参考图像,本文方法的评价结果要优于传统的MSE,PSNR以及SSIM方法,同时因为不变矩计算适用于不同尺度和变形图像,所以该方法可以对与参考图像尺寸不相等和各类变形的待测图像质量进行评价,并且计算简单。最后通过各类实测图像评价验证了方法的有效性。To solve the problems in the image quality assessment,taking the algorithm adaptability for images in different size and deformation into consideration,this paper proposes an image assessment method based on similarity of Hu invariant moments.By calculating 7 invariant moment values of the image,the distance of the invariant moments between two images can be calculated accordingly.Making use of the distance of the invariant moments to measure the similarity of two images,the quality assessment of the evaluated image can be done.The experimental result shows that the assessment result using this method is better than that using the traditional MSE,PSNR and SSIM methods for the evaluated image and the reference image in same size.At the same time,because the invariant moment calculation can be used for images in different size and deformation,this method can assess all evaluated images in different deformations with different sizes from the reference images easily.At last,the assessment for different actual images certifies the effectiveness of this method.

关 键 词：不变矩相似度 图像评价 图像大小 变形 

分 类 号：TP391.41[自动化与计算机技术—计算机应用技术]