| Three-dimensional(3D)video can provide stereoperception and realistic feeling of scene to audience,especially 3D-TV and Free-views TV,greatly to meet the demand of our life.The multiview video plus depth format is the most promising 3D video format.However,on the one hand,the blocking artifacts will be caused by 3D video compression,influences the perceived quality of 3D video;on the other hand,3D video quality assessment is one of the key technologies in 3D video system.Therefore,in this thesis,we study how to enhance the quality of 3D video while considering the relationship of color and depth information,and build 3D video quality assessment models which conform to human visual system through studying the factors that affect 3D video perceptual quality,then two effective 3D video quality assessment methods are proposed.The main contents are as following:Firstly,to eliminate the blocking artifacts caused by 3D image compression,a deblocking method is proposed based on joint dictionary.The proposed method mainly includes three stages: dictionary training,dictionary jointing and deblocking.At the dictionary training stage,an overcomplete color dictionary and depth dictionary are trained respectively with state-of-the-art dictionary learning method.At the dictionary jointing stage,for a testing sample,its corresponding color-depth joint dictionary is constructed based on the sparse coefficients with respect to the learnt color and depth dictionaries.At the deblocking stage,by estimating the reconstruction error threshold,a deblocking operation is performed to get the reconstructed 3D images with respect to the learnt joint dictionary.Experimental results demonstrate that the proposed method can effectively reduce the blocking artifacts of the compressed 3D images and generate high-quality synthesized images.Secondly,we studied the subjective perception quality assessment of 3D video.We designed and implemented subjective perception experiments through considering the influence of color and depth information on the perceived quality of 3D rendering views,and established a perception qualtity assessment database of 3D rendering videos.Which provides a data reference for designing quality assessment models of 3D rendering views.Thirdly,as existing 3D video quality assessment methods without fully considering the character of depth perception and the influence of view domain information,a reduced-reference quality assessment method is proposed based on DCT coefficients reorganization.The proposed method mainly includes three stages: 4D data set constructing,DCT transforming and reorganizing and reduced-reference quality assessing.At the 4D data set constructing stage,4D data set contained comprehensive information of 3D videos in spatial domain,view domain and temporal domain is constructed.At the DCT transforming and reorganizing stage,the data set is firstly transformed by 4D-DCT,and reorganized coefficients set is obtained by choosing complex DCT coefficients and reorganizing them.At the reduced-reference quality assessment stage,each subband in the reorganized DCT coefficients set is modelled by using generalized Gaussian distribution,and coefficients distance function and frequency ratio distance function are used to evaluate 3D video quality.Experimental results show that,the proposed method can achieve higher consistency with subjective assessment of 3D videos.Thus,the proposed algorithm conforms to human visual system.Finally,as the existing quality assessment methods usually measure the perceptual quality of the synthesized video in 3D system.a high efficiency view synthesis quality prediction(HEVSQP)metric for view synthesis is proposed.Based on the derived VSQP model that quantifies the influences of color and depth distortions and their interactions in determining the perceptual quality of 3D synthesized video,color-involved VSQP(CI-VSQP)and depth-involved VSQP(DI-VSQP)indexes are predicted respectively,and are combined to yield a HEVSQP index.Experimental results on our constructed NBU-3D Synthesized Video Quality Database demonstrate that,the proposed HEVSQP has a good performance evaluated on the entire synthesized video quality database,compared with other FR and no-reference video quality assessment(VQA)metrics. |
