Image Processing And Data Encoding System For Visual Prosthesis Based On ARM Cortex-A9 Platform

Vision is the most important means for human beings to obtain information from the outside world. 70% of the information accessed by human comes from the vision. Diseases like Age Related Macular Degeneration, AMD) and the Retinitis Pigmentosa(RP) could lead to vision function impairment, which causes a lot of trouble to the patients’ life. In recent years, using the visual prosthesis to generate functional electrical stimulation to help the patients restore the sense of vision has become a hotspot in this research area.Designing portable image processing and encoding system is an important part for the research and development of visual prosthesis. This system uses image processing algorithms to process the image data coming from the image acquisition system to the form that match the in vivo stimulator array. Comparing with ordinary image processing system, visual prosthesis image processing and encoding system has different requirements for algorithms and hardware equipment. Because of the limited number of stimulus electrodes, the information provided by the visual prosthesis is restricted. So the visual prosthesis system requires appropriate image processing algorithms to extract useful information from the original image efficiently. For hardware equipment, modules accelerating the computation are needed to acquire changing image information in the environment in time, and make the whole image processing procedure running in real-time. The system must have small scales and reasonable power consumption to make it portable. As the kernel unit for the visual prosthesis outside components, the image processing and encoding system also need abundant interface to connect other component like the camera and RF(Radio Frequency) module, and control them to work properly. It should be in charge of the power management and the interaction between the machine and researchers or the patients.This article introduces the current research of the image processing and encoding system in different research teams around the world. Their choice of hardware platforms and image processing procedures are analyzed, and the advantages and disadvantages of each choice are discussed. To accomplish the requirement of the visual prosthesis, a hardware platform using ARM Cortex-A9 processor is chosen for this research. Comparing with other similar platforms, this one has powerful processing ability, small scale, little power consumption, abundant interfaces and satisfying ability of controlling.With this platform, tow image processing strategies are implemented. The first one is for wide vision angle. It can obtain the information of the whole image scene. The second one is based on extracting region of interest which can get the detailed image information in a complex scene. Combining these two strategies can meet different vision requirement in everyday life. Software optimizations like coding with native code and using hardware acceleration are made to reduce the processing time. After the optimization, an image with a resolution of 640*480 can be processed to targeted resolution of 32*32 in real-time. The stimuli mode of the visual prosthesis is studied, and stimulus strategy based on the requirement of the stimulator is designed.