| With the civilization and progress of mankind, more and more intelligent science was applied to the reality of society. Intelligent science as the intelligent simulation science, the development of information technology that becomes more mature makes more and more applications products. Human visual information processing has evolved to a high degree of perfection stage.A large amount of information that the brain receives from the outside world, the vast majority is dealt with the visual processing system. Visual information processing mechanism is one of the major issues in psychology, neuroscience, computer science and others. The human visual system is one of the most comprehensive known and complex mechanisms of information processing system. Its study is of great significance with the understanding the ability and mechanism of human information processing and relations between information expressed in the brain and the material world. On the other hand, with the rapid development of computer science, to create an intelligent computer like human brain is one of the greatest expectations of the researchers. To make this expectation a reality, we must first make a computer have capacity like brain that deal with a multitude of complicated information (particularly the visual information), but this major issue although the computer scientists decades of efforts have yet to be resolved. The most one of the reasons is that we still do not have a thorough understanding of the human visual system work mechanism.In the visual physiological theory, extraction step by step and parallel processing mechanism of visual information is more and more physiological experiments proved. Natural images from the outside world through the photoelectric conversion process of the retina are from the optical signal into the nerve signal that the nervous system can deal with. The neural signal as the carrier of input visual information is delivered primary visual cortex. Nerve cells processing this visual information provide information infrastructure for senior intelligent. Transmission processing of visual information is known as visual pathway in the visual physiological. After visual information is processed, it is transformed into hypothalamus through the optic nerve and fiber optic tract. In the hypothalamus it is divided into two channels: (1) LGN - cortical pathway is a key pathway visual imported. (2) Superimposed body - hypothalamus pillow - cortical pathway, only 10% of primate retina through this fiber pathway linked with the establishment of the visual cortex.Area 17 in the visual cortex is known as the first district (V1) or striate cortex. It accepts the direct input of the LGN, so known as the primary visual cortex. The research for the function of the visual cortex is almost on this cortex. This is the visual cortex handle the starting point of information. Starting from the primary visual cortex, Visual information through a number of channels is transformed into more advanced visual cortical processing. V1 mainly includes two types of nerve cells: the simple cells with towards selective and the complex cells that are not sensitive to the position.Based on the physiological mechanisms for dealing with visual information, this paper presents a multi-layered visual features information processing model, which describes the processing of the visual information from natural Image boost to senior-level information. Model is divided into three main processing, primary visual cortex, the symptoms network layer, semantic network layer. Primary visual cortex is the bottom; it is the basis of the network layer. Primary visual cortex includes the physiological function of complex cells, the retina, the LGN and simple cells. Shape is an important part of the visual content of images. It also is basis of the high-level cognitive thinking image. Shape information in natural image is composed of the edge with different directions. So edge detection is the basis of shape feature detection. From the perspective of gray level image, edge is the place gray changes. Edge also is the outline information of objects. Whether the image understanding, pattern matching, edge contour information is an important basis for judgment. Based on simple visual cortex cells and cell complex visual information processing features and functional linkages between them, this paper presents the basic steps of visual information shape detection, and is divided into the edge detection and the basic shape detection.Edge detection is the key information processing in areas of image processing technology. At present there are already some handling operators. For example, based on partial differential image of the traditional edge detection technology Sobel operator, Canny operator, and so on. According to the needs of multi-level visual features information processing model, the paper analysis of traditional edge detection operator. Combining advantages of them, an edge detection model based on BP is designed. The model uses artificial neural network features, and combines the advantages of traditional operator. The model detects the image edge.The paper made a lot of experimental analysis on Canny operator, Roberts operator, Sobel operator, the operator and Log Prewitt operator, and made a lot of experimental analysis on the edge detection model based on BP.The paper uses Lenna image that is the most commonly used in the field of image processing. Experiments simulate the visual information processing through the edge detection model based on BP.Given the limited level of mine, some still need to improve on. For multi-level visual information processing model, I have the following outlook:1. Noise and the marginal position is difficult point in edge detection. Pseudo-edge direct impact on the application of information technology based on the edge characteristics. Edge detection algorithm in the treatment is based on the pixel level. When the picture at a higher resolution, detection algorithm efficiency will be significantly affected. How to solve these problems will be worth further study.2. On the issue in object recognition, visual information processing is the first large fine after. That is the first major study on the contours, then the need to study the details. The contour information of large-scale perspective play a key role in the fields of object recognition, image understanding, pattern recognition, and so on. How to solve the shape information accurately detect effective expressions is the key issue of the representation net layer, also is follow-up work of this paper.The prospect of Visual information feature detection is optimistic. Our research work will also be endless. We are standing on the shoulders of our predecessors, and are based on now, focus on the future. |
