Optimization Of Simulated Prosthetic Vision Based On Global Contrast Saliency Model

Visual prostheses provide an effective solution to recover visual function in the world and bring hope to the patients who suffer from retinitis pigmentosa and age-related macular degeneration.By using an electrode array to electrically activate surviving nervous tissue,visual perception can be stably evoked.Recently,visual prostheses have made encouraging progress,some products have received EMA and FDA approvals for the treatment of retinitis pigmentosa.However,due to restrictions of electrode manufacturing,biological and physical,visual prostheses' electrode number is still low and it can only generate low-resolution visual percepts.What's more,the researchers found that phosphene has irregular shape,color,existing distortion and dropout phenomena.Clinical reports demonstrated that currently the implanter' visual acuity is very low,it is difficult for them to complete daily tasks.In order to provide more useful information to implanter by using the limited electrodes,and improve their ability to conduct visual tasks in daily life,we developed the research of optimization of simulated prosthetic vision based on global contrast saliency model.To improve the ability of object recognition,we proposed to utilize GS_GD saliency model,morphological reconstruction and connected-region-select methods to detect foreground object.After that,edge information was combined with foreground information to enhance the image.Psychophysical experiments verified that the method significantly improved the subjects' performance of object recognition and the strategy is effective.At present,the algorithms used for optimizing visual percepts are complex and need large calculation,visual information cannot be processed in real-time.In order to solve the problems,image color contrast and intensity difference are used to extract the saliency object in this research.Then we use edge information to highlight the foreground and restore background information.Eye-hand coordination experiment results showed that our strategy can optimize visual information in real-time under low-resolution condition and prominently improved the performance of subjects.The researches in this paper effectively optimized visual percepts under low-resolution condition and improved subjects' ability to complete daily tasks.It is hoped to provide theory evidence for the design of image process module of visual prostheses in the future.