基于灰度共生矩阵相似图的图像质量评价方法  

Image quality assessment method based on similarity maps of gray level co-occurrence matrix

作　　者：孙荣荣[1] SUN Rongrong(Shanghai Institute of Measurement and Testing Technology,Shanghai 201203,China)

机构地区：[1]上海市计量测试技术研究院,上海201203

出　　处：《计算机应用》2020年第S01期177-179,共3页journal of Computer Applications

摘　　要：针对图像质量评价(IQA)问题,提出一种基于灰度共生矩阵相似图(GLCMS)的方法。首先,分别得到参考图像和失真图像的灰度共生矩阵(GLCM);然后,求得此两幅灰度共生矩阵的相似图,并提取相似图的标准差和熵作为失真图像的特征向量;最后,将特征向量输入到支持向量回归(SVR)算法预测图像质量。TID数据库是专为评价图像质量而建立的,TID数据库上的实验结果表明,所提方法无论在训练集还是测试集上,与主观评价方法的斯皮尔曼相关系数和皮尔逊相关系数均达到0.93以上,表明此方法较好地符合人类视觉特性。该方法为图像质量评价方法提供了新的思路,可用于图像质量评价,使其与人类视觉具有更好的一致性。In order to solve the Image Quality Assessment(IQA)problem,a method based on Similarity maps of Gray Level Co-occurrence Matrix(GLCMS) was proposed innovatively here. Firstly,the Gray Level Co-occurrence Matrixes(GLCM)of reference image and distorted image were obtained. Then,the similarity map of the two GLCM was calculated,and the standard deviation and entropy of the GLCMS were extracted as the feature vectors of the distorted image. Finally,these feature vectors were taken as the input to the Support Vector Regression(SVR)algorithm to predict the image quality. The TID database was established to evaluate image quality specially. The experimental results on the TID database demonstrate that the Spearman and Pearson correlation coefficients of this method with subjective method are all above 0. 93,no matter in the training dataset or in the testing dataset,it shows the proposed method fits well with human visual characteristics. This proposed method provides a new way of thinking for IQA,and achieves better subjective perceived consistency.

关 键 词：图像质量评价 灰度共生矩阵 相似图 标准差 熵 支持向量回归 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]