Research On Visual Attention Mechanism And Its Applications

As one of the important characteristics of the human visual system, visual attention mechanism enables the human to quickly select from a complex scene object to a small number of significant concern. It is an important means for human beings with limited resources to handle large amounts of information. In view of its efficient data filtering capabilities, the introduction of the visual attention mechanism into the field of computer information processing, in particular the image processing field with the need of calculating mass data, is of great theoretical and practical significance. Therefore, the integration of the physical anatomy, cognitive psychology and computer science and other disciplines of visual attention to the results of the calculation model quickly became the image processing research focus.This paper focuses on bottom-up model of visual selective attention mechanism through theoretical study and experimental analysis. This paper introduces the mechanism of visual attention and its present, and described the mechanisms of visual attention based on physiological psychology, followed by a detailed study from the static and dynamic aspects of computational modeling of visual attention.At the static aspect, a detailed study and implementation of Laurent Itti et al proposed for still image model of visual attention is achieved. Conducting a large number of experiments that Laurent Itti model based on the extraction of color features in the significant when the possible phenomenon of regional inversion, analyze their causes and proposed improvements. Meanwhile, in order to better highlight the significant areas of significant maps of the merge strategy were optimized. Comparative experimental results demonstrate the improved effectiveness of the program.In addition, the project also studied the movement of significant model calculation, a calculation is given the contents of each frame image sequence was significant movement of the program. First, extract the current frame by block matching motion vector and breaking it down into four directions of the scalar, the parties are calculated significant upward movement, and finally fused into motion saliency map. Stationary camera image sequences collected experimental results show that the results of the program results with basically the same perception of human eyes to verify its validity.To apply the visual attention theory, based on this assumption that the detection goal is significant, the visual attention mechanism is introduced into the image matching field and the whole process is divided into two stages of preprocessing and matching. At preprocessing stage, a saliency map of the image to be matched is produced according to visual attention model; at matching stage, the salient region is matched with the template image in descending order of significance. Experiments show that the method of introducing the visual attention mechanism in the matching stage can detect the target more quickly.