Research On Evaluation Model Of Human Vision In Stereo Video Based On Multi-scale Analysis And Similarity

Stereo video quality assessment is the basement of perception and display and also the foundation of stereo video system design. The subjective methods for stereo video quality are consistent with the results of human subjective perception, but because of its poor stability, poor portability, time-consuming, costly, it will be greatly limited in practice. The existing objective evaluation methods for stereo video quality mainly considered the extent of the loss of image information, and little concern the subjective perception of stereo images, therefore they can not reflect the merits of video quality. With the continuous development of stereo video technology, how to effectively evaluate the stereo video quality will be a hotspot and difficult.Firstly, we simply introduce the subjective and objective methods of the stereo video quality, and give an analysis of several typical objective evaluation models; Then, we analysis in detail the basic characteristics of human visual system and wavelet transform theory. Since human being is final receiver of the image, the key point of the merits of the video assessment is that it should match the characteristics of human visual system (HVS). A method for stereo image quality assessment based on multi-scale analysis and similarity is proposed based on the basic HVS characteristics such as visual non-linearity, contrast sensitivity, multi-channel model and masking effect and so on. The algorithms are described specifically as follows:1.The objective evaluation method of the stereo video based on the characteristics of HVS. The method is based on the difference information between stereo images. The algorithm to the evaluation of stereo video quality is divided into two aspects: the stereo video quality assessment and the stereo sense quality assessment. Viewpoints of each frame stereo video sequence are firstly performed five level wavelet decomposition, the whole spatial frequency of image is divided into six parts, then calculate the similarity index of each frequency sub-band; According to the contrast sensitivity characteristics, the quality evaluation results of each band is weighted average, to get the final stereo video quality assessment index by fusing of the index of left and right view; Secondly, the difference image through multi-scale analysis and similarity evaluation, obtain the assessment index of stereo sense.2.The objective evaluation method of the stereo video based on the human depth perception. The method is based on the disparity information between stereo images. Extracting information of disparity use a one-dimensional window stereo matching method, it can limit the scope of matching to the stereo images of the corresponding horizontal scan lines, and the left and right disparity information can access simultaneously through a single scan, which greatly reduces the computation of matching and improve the matching efficiency. After the obtained disparity map by the multi-scale analysis, we calculate the correlation scale of wavelet coefficients to obtain the evaluation index of the disparity information. Finally, several stereo video quality assessment measures are obtained by the linear fusion of the disparity information evaluation index and stereo video quality assessment index.The experimental results show that the proposed stereo video quality assessment algorithm based on the multi-scale analysis and similarity in this paper easy to implement and fast. The algorithm that taking full account of visual perception, the relevance of the proposed objective evaluation index and the human subjective evaluation value outperforms traditional PSNR method, has the higher consistency with the human subjective feelings, and it may reflect the merits of stereo video quality and stereo sense.