Objective Image Quality Assessment Based On Image Structure Information And Natural Scene Statistics

With the rapid development of Internet technology and the popularity of multimedia content in image and video,in order to meet people's visual enjoyment,we need to continuously improve the performance of the image processing system.Image quality is an important index to measure the performance of the image processing system and to optimize the parameters of the image processing system.In this paper,three image quality evaluation models are proposed,and they are in good agreement with human subjective feelings.Image quality assessment for multiply-distorted images is the emphasis and difficulty in image quality assessment(IQA)filed.Based on high-order phase congruency,a no-reference IQA method for multiply-distorted images is proposed.Firstly,the high-order phase congruency is computed to capture the structural information of the image.The statistical features of four orders phase congruency are extracted by gray level co-occurrence matrix(GLCM),respectively.Secondly,based on the analysis of the correlation between adjacent orders of phase congruency and the correlation between adjacent orders of local entropy of phase congruency,the mutual information and cross entropy of that are calculated.Finally,the support vector regression is utilized to build a regression model and then it is used for quality predicting.Blind image quality assessment(BIQA)assesses the perceptual quality of the distorted image without any information about its original reference image.Features,in consistent with human visual system(HVS),have been proved effective for BIQA.Motivated by this,we propose a novel general purpose BIQA approach.Firstly,considering that HVS is sensitive to image texture and edge,the image gradient and wavelet decomposition is computed.Secondly,taking the direction sensitivity of HVS into account,the gray level co-occurrence matrixes(GLCMs)are calculated in two directions at four scales on the computed feature maps,i.e.,gradient and wavelet decomposition maps,as well as the image itself.Then,four features are extracted for each of GLCM matrix.Finally,a regression model is established to map image features to subjective opinion scores.In recent years,no reference image quality assessment has attracted more and more attention.In this paper,we propose a novel no-reference image quality assessment algorithm based on Gray Level Co-occurrence Matrix(GLCM).Firstly,the image is decomposed by wavelet transform,meanwhile,the Local Binary Pattern(LBP)feature of the image is extracted.Second,the GLCM of the wavelet subbands and LBP map is calculated.Four feature parameters of the GLCM are taken as the feature vector of the image.Next then,the mapping relation model between the feature vectors and subjective scores is obtained using the support vector regression(SVR).Finally,the model is utilized to predict image quality.Experimental results demonstrate that the proposed method performance outperforms the selected image quality assessment methods,and the predicted score is well consistent with the subjective assessment result,thus the method can reflect the human visual perception of image quality effectively.