An Expert System For Television Picture Quality Evaluation

As the information revolution swept the world and the rapid development of information technology, a deep digital revolution is also taking place in the field of radio and television. The transition from analog TV to digital TV has begun. Currently, high-definition digital TV has penetrated into all aspects of people's daily work, life and entertainment, and people's requirements for the quality of video are getting higher and higher.At present, the picture quality of most televisions on the market is not consistent, and the reason is that the picture quality of television needs to be evaluated many times in the design and production process. The limitations of existing methods may result in different results of evaluation, and thus leads to the less consistency of the television picture quality.Traditional television picture quality evaluation methods-subjective assessment methods suffered from disadvantages of being time-consuming and costly, and the external environment and human factors may impacted on the results. In this paper, an objective evaluation method based on feature extraction was proposed, and the major innovation of this method was to transform the subjective indicators for subjective evaluation to objective indicators. Specifically, we extracted features which can reflect the picture quality from the images of the television using image processing methods, and then selected a few features of good approximation to simulate the subjective indicators. Finally, an expert system based on support vector machines (SVM) was established, which can evaluate the television picture quality intelligently.In feature extraction stage, we extracted multiple features, and selected the optimal combination of these features which can reflect the trends of the changes of subjective indicators through a large number of experiments.In the design of expert system, we chose the SVM as our classifier based on the analysis of the advantages and disadvantages of commonly used classification algorithms. SVM was proposed for the learning and classification problems with small samples. It could overcome the over-fitting problem and have strong capability for nonlinear classification. In addition, a multi-classifier based on binary tree model was proposed, which can overcome the shortcomings of traditional methods and improve the efficiency and accuracy of classification.The experiment results showed that the method proposed in this paper could evaluate the picture quality quickly, and the evaluation results were consistent with those obtained by the subjective evaluation method, which proved the accuracy and feasibility of our method.