Research On The Model Of Perception Quality Prediction For 3D Video Terminal

With the rapid development of 4G generation Communication, digital communication has gradually become the dominant multimedia communication. While for the imperfect basic communication infrastructure, the image may well be distorted, in terms of its distortion type and degree, during the process of image’s capture, processing, storage and transmission, greatly affecting the ultimate perception quality for viewers. In this paper, we fully analyzed the fundamental characteristics of human vision system, based on which three effective approaches are put forward:In the first place, we introduced the relationship between the human vision system and human perception so as to deepen the root of feature derivation; then we introduce the predominant methods of image quality assessment, particular reduced reference image quality assessment indice. Finally three effective reduced reference image quality approaches are put forward:1) A reduced reference image quality assessment approach based on energy change in reorganized domain; As far as the traditional IQA is concerned, the pixel based IQAs fail to fully consider the characteristics of human vision system such PSNR SSIM in spite of its high efficiency. The proposed method extracted energy change in reorganized domain on the basis of multichannel characteristics of human’s eyes, and the change of the energy between distorted and referenced images indicators the perception quality. The experiment shows the proposed IQA has high efficiency on causes of its computation complexity and prediction accuracy.2) A reduced reference image quality assessment approach on Wavelet transform for Gaussian blur distorted image, the energy information is extracted in the wavelet domain firstly and a strategy of measuring the distance between the reference and distorted images is formulated, Experimental result shows that the proposed metric earns high consistency with the subjective quality and requires low computation complexity practically.3) A reduced reference image quality based on statistical characteristics and energy, the algorithm extracts the shape parameter of its reorganized DCT transform which conforms with the general Gaussian distribution, and energy is derived from the image’s reorganized DCT domain. Finally, the support vector regression is utilized to predict the perception quality by means of sample training and testing. Experiment results show that the predicted quality is highly consistently with the subjective quality. As a reduced reference IQA, it yields great prediction accuracy and very simple algorithm complexity.