Research On Infrared Image Enhancement Algoriths For Prosthetic Vision

As an effective solution to the recovery of visual function,visual prostheses have achieved outstanding results in the treatment of diseases such as age-related macular degeneration and retinitis pigmentosa.However,due to the restrictions of biocompatibility and electrode manufacturing,the implantable electrode number of the visual prostheses is still very limited.And the visual perception of the implants is still in low resolution,with distortion,dropout,irregular shapes and uncontrollable colors.Clinical research demonstrated that appropriate image processing strategies can effectively improve the visual perception of the implants.But it is impractical to carry out experiments on implants for expensive implantation cost,limited implantation number and so on.So researchers put forward simulated prosthetic vision(SPV)for the research of image processing strategies and made great progress.However,current researches are all based on visible images and visual prostheses fail to work at night without images captured by cameras.Infrared imaging technology,which has been widely used in military,surveillance and other fields,manages to solve the problems at night.To improve the environment of visual prostheses,this study applied infrared imaging technology to visual prosthesis and explored proper enhancement algorithm for prosthetic vision.In this study,we use infrared camera to capture images at night.According to the characteristics of infrared images,we compared three representative infrared image enhancement algorithms.After qualitative and quantitative evaluation,we selected double platform histogram algorithm for SPV research.Results showed that double platform histogram enhancement doesn't have obvious enhancement effect under low-resolution prosthetic vision.To solve the problem,we put forward a double platform histogram algorithm based saliency detection and carried out several eye-hand coordination tasks.Research demonstrated that the accuracy of visual tasks reached 62.11% under the algorithm we proposed,increased by 13.93%.And the time to complete the task reduced from 69.26 s to 57.83 s.We hope the research could help expand the work environment of visual prostheses in the future.