Perceived Visual Field Expanding Based On Content-Aware Method Under Simulated Prosthetic Vision

Retinal prostheses have shown great value in restoring vision.Even though great progress has been made in the development of retinal prostheses,small visual field visual percepts still remain as major obstacle for implants to achieve satisfying prosthetic vision.Hence,we aim to propose a novel method to expand the small visual field by image processing approaches,so as to achieve higher performance of prosthetic vision.This study focused on expanding perceived visual field using image retargeting techniques under simulated prosthetic vision(SPV),compressing an image of a wider visual range into the small visual field of prosthetic vision.Based on the content-aware image retargeting method,the importance map was calculated from the two image features including gradient and color,according to which each pixel is scaled at different degree,resulting in more important content preserved.We compared our method with cropping,scaling and seam-assisted shrinkability(SAS)method under SPV by psychophysical experiments.Results showed that our method outperformed in preserving more key features and had significantly higher recognition accuracy in comparison with other three image retargeting techniques under the condition of small visual field and low-resolution.Our improvements on the content-aware image retargeting strategies was verified to be beneficial to transfer more useful visual information and improve the performance of prosthetic vision under simulated lowresolution prosthetic vision.We hope they contribute to the development of visual prostheses,providing more benefit for prosthesis recipients in their daily life.