Saliency-based image processing for retinal prostheses

We present a computationally efficient model for detecting salient regions in an image frame. The model when implemented on a portable, wearable system can be used in conjunction with a retinal prosthesis, to identify important objects that a retinal prosthesis patient may not be able to see due to implant limitations. The model is based on an earlier saliency detection model but has a reduced number of parallel streams. Results of a comparison between the areas detected as salient by the algorithm and areas gazed at by human subjects in a set of images show a correspondence which is greater than what would be expected by chance. Initial results for a comparison of the execution speed of the two algorithm models for each frame on the TMS320 DM642 Texas Instruments Digital Signal Processor suggest that the proposed model is approximately ten times faster than the original saliency model.