Video Quality Assessment Models With Human Visual System

Technology used to assess video quality automatically plays significant places in today's video processing areas. Visual information is the most important ways for human to evaluate information. With the rapid progress of personal computers, digital communications, multimedia and network technology, digital video and image information are becoming one of the most important media, which has been deep into people's daily lives. Further study of digital video quality evaluation method is necessary, it can test the evaluation system design advantages and disadvantages of the structure and algorithm, the optimization of the system at the same time have an important guiding role.This article in the current field of video quality assessment work, a combination of random neural network with the new method of motion vector, and the use of its degradation of human visual perception is estimated. The results are processed with an effective non-linear fitting, and have well correlations with human visual experience. At the same time, compared to other methods, this paper proposes a more universal approach and accuracy. This paper includes the following aspects:1. The introduction of the HVS (Human Visual System / HVS) model to determine the optical properties of images based on visual perception of image quality evaluation method, that is based on perceptual characteristics of the human visual to light the extent of evaluation of image quality, more in-depth on the HVS study to address the objective evaluation of the visual model and the subjective visual effect of the bias, so that an objective evaluation correlations to subjective feeling of human visual feeling. Introduced the study and then used the neural network model, cellular neural networks features the most attractive and the human visual system has lots of similarity. Through feature extraction and neural network simulation to resolving the reference in some measure in the objective evaluation of the problem of a narrow scope of application.2. This paper promotes a full reference video quality evaluation method correlates to human visual perception. The combination of human perceptual characteristics, time and space from the video feature extraction from, a measure based on the importance of the concept of perceived and used in an objective evaluation of video quality. By reference to the video feature extraction algorithms used in accordance with the importance of different video information and the importance of perceptual measurement, the final result of distortion and reference video of the differences in perception of information access to video distortion and quality metrics.3. the implement of a vision based on objective characteristics of video quality evaluation model using the random neural network training methods, with the parameters of simple, efficient performance characteristics. The software bases on Windows operating environment and the preparation of the VC/Matlab system. The experiment uses VQEG Phase I video data and the subjective DMOS data points, to compare the results of a variety of real-time applications. Be compared to pixel-based statistical algorithm such as an extension of PSNR, the proposed RNN algorithm has good and moderate complexity,...