The Influence of Gray Scale on Image Recognition Under Simulated Prosthetic Vision

Retinal prostheses have become an effective means to help restore functional vision in patients with retinal degenerative diseases. One of the basic needs of prosthesis implant recipients is the ability of recognizing objects in daily life. However, due to the limited number of stimulating electrodes, retinal prostheses can only produce a low resolution perception. Therefore it is an important task to study and apply different image processing methods to optimize the information presented to the prosthesis wearer. In order to find the basic factors affecting the recognition of objects and scenes in daily life, this paper carried out the study on the influence of gray scale on image recognition under simulated prosthesis vision. Based on the image processing strategy of foreground extraction, the image recognition with and without gray scale for two image processing methods at five resolutions (16×16, 24×24, 32×32, 48×48, 64×64) were investigated. The results of psychophysical experiments proved that the presence or absence of gray scale has no significant effect on image recognition at 16×16 and 24×24 resolutions, and when the resolution gradually increased from 32×32 to 64×64, the images with preserved gray scale could achieve higher recognition accuracy and a significant difference between the two recognition results. Retaining the gray scale of images has prominent advantages in complex scene perception, especially when overlapping occurs in the scene.