An Image Strategy Based on Saliency Detection Using Luminance Contrast for Artificial Vision with Retinal Prosthesis

Due to the limited number of implantable microelectrodes, subjects worn by retinal prosthesis received only limited discrete light spot (called artificial vision). Thus, the external abundant information is largely lost. Whereas there is little room for significant improvement in the electrodes number, it would be a feasible way to optimize the low-resolution information in the prosthetic vision by effective image processing. Based on luminance contrast, a real-time saliency detection strategy was presented to enhance the artificial vision, which included color space conversion and a visual attention simulation processing method. The strategy was evaluated by two benchmark databases, and its superiority was validated in extracting image foreground compared with the basic luminance contrast and other algorithms. Meanwhile, simulated experiments were conducted to verify the proposed strategy in the retinal prosthesis.