Based On The Calculation Model Of The Image Recognition Mechanism Of Classical Receptive Field

Computer vision is an important aspect in the research of intelligent computers. However, most of the researchers focus on theories and algorithms proposed from the standpoint of engineering. The underlying physiological principles are ignored and the theories lack a complete computational system and uniform algorithms, which makes it impossible to solve problems of complex scene recognition. Therefore, contemporary computer vision processes only localized information yet most natural scenes are composed of multiple different local information. By contrast, vision perception system of higher mammalian animals is capable of perception tasks in various complex environments. With local visual information well combined in the brain, mammalian animals produce aggregated and harmonized sensation when observing scenes. They are able to distinguish objects of different sizes and features from highly complex background. The research of visual perception of higher mammals hence makes a possible break through. Neural physiological theories and psychological theories will provide new prototypes in computer vision, which help to conquer difficulties currently encountered.In the visual system, visual signal is first primarily processed in the retina (also known as peripheral brain). Ganglion cells are in the last phase of retinal information processing. Receptive field is the primary structure and functional unit in visual information processing. The output signal and response features of the complicated network in the retina are reflected ultimately by the receptive field of ganglion cells. It is where low level information converges and it provides input to higher processing stages. Its foundational role reveals the importance of research into the information processing mechanism of retinal ganglion cells.The classic receptive field of retinal ganglion cells is concentric. According to its spatial summation features, it processes brightness and contrast of images and it is sensitive to edges in the images. Yet the complex visual system of higher mammals is definitely not to be limited to edge detection. In addition to edge detection, it passes much more information to the brain.The non-classical receptive field (nCRF) is a wide range of areas outside the classical receptive field (CRF). Stimulating this region alone does not cause cell responses directly, but it can adjust the response induced by the stimulation inside CRF. The non-classical receptive field of retinal ganglion cells plays a role of disinhibition, so it can, to some extent, compensate for the loss of low spatial frequency information caused by the classical receptive field. It transmits the brightness gradient of the image while maintaining border, showing the slow changes in brightness on the surface of the large area. Thus, the non-classical receptive field greatly broadens the information processing scope of the visual cells, providing a neural basis for integration and detecting a wide range of complex graphics. Study finds that the receptive field (RF) of retinal ganglion cells (GCs) changes with the different visual stimuli, but previous modeling of retinal ganglion cells in non-classical receptive field was mostly based on fixed receptive field, ignoring the dynamic changes in receptive field properties.The aim of this paper was to construct a bio-inspired hierarchical neural network that could accurately represent visual images and facilitate follow-up processing. Our computational model adopted a ganglion cell (GC) mechanism with a receptive field that dynamically self-adjusts according to the characteristics of an input image. For each GC, a micro neural circuit and a reverse control circuit were developed to self-adaptively resize the receptive field. An array was also designed to imitate the layer of GCs that perform image representation. Results revealed that this GC array can represent images from the external environment with a low processing cost, and this non-classical receptive field mechanism could substantially improve both segmentation and integration processing. This model enables automatic extraction of blocks from images, which makes multi-scale representation feasible. Importantly, once an original pixel-level image was reorganized into a GC-array, semantic-level features emerged. Because GCs, like symbols, are discrete and separable, this GC-grained compact representation is open to operations that can manipulate images partially and selectively. Thus, the GC-array model provides a basic infrastructure and allows for high-level image processing.This paper constructs a multi-layer network computing model to make general representation experiment on the natural image. The results show that the similar areas which are adjacent to each other would be gathered together, representing by large receptive fields. While those are not similar to each other would be separated, represented by several some receptive fields. Compared with other original methods, this method has higher speed, efficiency and better performance. In addition, In addition, the clustering and outline fitting experiments on represectation image, reach very good results. Therefore, the multi-layer network computing model indeed provides a good foundation for subsequent processing (e.g., segmentation, etc.) provides, also helpful to the understanding of the image.