The Research Of Image Processing Algorithm For Visual Prosthesis

The World health organization states that there have about 45 millions blind persons in the world with the annual increase of 7 millions. Even the low vision population is up to 170 million, and nearly about 80% of them can be treated.. With the development of modern electronics, biology and the information of science,people can make the blind regain the ability of visual sense by the visual prosthesis technology.Image processing system is indispensable parts of the visual prosthesis, which includes that the camera captures images and that it uses a series of typical image processing algorithms to extract image useful characteristics information.Aimed at the deficiencies in visual prosthesis image processing, this thesis present two improved:(1) For complex scene image, the edge character extracted by the traditional image processing includes a number of background and texture, while these Characters have no significance to identifying objects, and produce the burden of visual information transmission on the micro-based electrode stimulation. In consequence, this thesis puts forward of the focus area detection of the complex image based on the visual saliency.The sub-image centered on the detected ‘focus’ then can be used to the traditional image processing method for edge extraction.(2) In the visual prosthesis image processing, the traditional edge detection method is based on single scale differential operator. The scale section is very important for the edge detection: small scale operator has less noise robustness but the high edge location precision; the small scale operator is on the contrary. Combined with multiple scale theory, this thesis presents a multi-scale edge detection method,which fuses soblel edge strength of several scales to get the final edge strength map.The fused edge strength inherits the advantages of different scales and achieves overall better performance. Meanwhile the threshold for edge detection is adaptively using the Maximum between-class variance and it is applicable to edge detection of the more complex image.Appling two improved points to the visual prosthesis processing can obtain the satisfactory result of image processing. This has practical significance to improve vision restoration system.