Research On Motion-based Visual Attention Computational Model

Visual attention mechanism is an important psychological mechanism of human visual system. Using visual attention mechanism, people can extract useful information quickly, response to the environment timely and effectively. Based the mechanisms of the human visual system,the computational model of visual attention is a hot field of image processing research, the introduction of visual attention mechanisms contribute to rational allocation of image processing resources, improve image processing efficiency, and has a broad application prospect.This paper is mainly based on bottom-up model of visual attention, thesis work includes two parts of the static model and dynamic model.Paper's study on the static model bases on Itti model, by analysis of structures and the key algorithm of computational model of visual attention, we improve the algorithm on the color and intensity channels, experiments show that the improved algorithm is more consistent with the characteristics of the human visual system, and is conducive to the model top-down information to add; the same time, the direction of features;Meanwhile, by analyzing the characteristics of Gabor filter, we improve the characteristics of orientation channel of the model.Traditional Itti model mainly concentrate on the calculation of static image region, and so visual attention model based on motion features is conducive to visual attention model's expansion to the dynamic scene, which can make the model more similar to the biological visual system.This paper proposes a new algorithm which is verified by experiments to add motion feature to the model.Thesis makes a complete analysis of the idea of visual attention and methods of computation, to achieve a good performance in the applicability of visual attention computational model,and improve the visual attention computational model's theoretical and practical value. Experimental results show that the proposed model could meet the law of visual cognition, and the model's results are in line with the basic characteristics of human visual perception.