Perceptual Full Reference Image Quality Assessment

With the fast development of the Internet,image data,which is featured by intuitionistic expression,rich content and easy to share,has gradually become a vital medium in the process of sharing knowledge,transmitting information and expressing emotion.The wide application of image data challenges various processing procedure in the image system,such as image storage,image transmission,image analysis and image display.In these procedures,images are affected by noise inevitably,which results in the decline of visual image information.Since human being is the ultimate perceiver of images,and the decline of image information would reduce the visual experience of human,image quality has become an essential standard in estimating image processing techniques.Evaluating the image quality automatically,effectively and accurately is of great guiding importance to image processing techniques.This thesis focuses on the analysis of the human visual system and full reference image quality assessment and its application.The main contribution of this thesis is four-fold:(1)We propose an image quality assessment method called image decomposition-based structural similarity index(IDSSIM).Perceptual image quality assessment adopts a computational model to assess the image quality in a fashion,which is consistent with human visual system.From the view of human visual system,different image regions have different importance.Based on this fact,we propose a simple and effective method based on the image decomposition for image quality assessment.In our method,we first divide an image into two components: edge component and texture component.To separate edge and texture components,we use the TV flow-based nonlinear diffusion method rather than the classic TV regularization methods,for highly effective computing.Different from the existing content-based IQA methods,we realize different methods on different components to compute image quality.More specifically,the luminance and contrast similarity are computed in texture component,while the structural similarity is computed in edge component.After obtaining the local quality map,we use texture component again as a weight function to derive a single quality score.Experimental results on five datasets show the effectiveness of the proposed methodthe.(2)We propose an image quality assessment method based on global statistics and structure information.Traditional full reference IQA approaches often consider that the reference image has perfect quality,ignoring the fact that some distortion types(e.g.contrast change)would enhance image quality in some cases.Therefore,it is difficult to evaluate the enhanced image quality by using the existing image quality assessment methods.In order to deal with this problem,the proposed method considing both global statistics and structure information of image.First,the low frequency entropy is used to calculate the global statistics similarity,and the primary structure weighting function is used to calculate structure similarity.Experimental results on four databases show that the proposed method achieves the state of the art performances.Moreover,the experimental results also show that the proposed method can effectively improve the evaluation accuracy for contrast change distortion type in the conditions of no influence for other distortion types.(3)We propose an image quality method called diffusion speed structure similarity(DSSIM).The purpose of the image quality assessment(IQA)is to evaluate image quality consistently with human subjective evaluation.In IQA,many structural features have been used to measure the quality degradation of an image.However,most of the existing IQA methods are patch-based,in which the quality evaluation is only performed within a patch independently,and the changes among neighboring patches are ignored.In fact,HVS perceives distortions not only depends on local structural(intra-patch)distortions,but also relates to the structural distortions of their neighborhood(inter-patch).However,the inter-patch distortion is not characterized in the traditional structural features.Based on this insight,we propose a new IQA method by considering the intra-and inter-patch correlations.Specifically,the image diffusion speed is used as a visual feature to characterize the interpatch information.In addition,the image gradient is used to measure the changes of the intra-patch structure.In the pooling stage,the image diffusion speed is also employed as a weighting function to reflect the importance of a local region.Experimental results on four databases demonstrate the validity of the proposed method..(4)We apply the image quality assessment to the QR code beautification and propose an image quality assessment based QR code beautification.Two stages are involved in this method.One is the visual saliency and halftoning based weighted stage,in this stage,the corresponding saliency image and halftone image are obtained by visual saliency detection and halftone techniques,then,the modules distribution of QR code is optimized by according to the saliency image and the halftone image.The other stage is the codeword distribution optimiztion and the evaluation score of image quality is used as the optimization standard during this phase.The experimental results show that the proposed method substantially enhances the appearance of the QR code.