Optimization Of Global Saliency Detection Based On Multi-Visual Features And Its Application In Embedded Platform

Age-related macular degeneration and retinitis pigmentosa are two incurable retinal degenerative diseases.The pathogenesis of both diseases is due to the progressive degradation of retinal photoreceptor cells,which ultimately leads to blindness.The development of visual prosthesis has made it possible for blind patients to reconstruct artificial vision.The visual prosthesis system usually consists of an image capturing device,an image processing unit,a radio frequency transmitting module,a signal receiving module and a stimulation electrode array.However,due to factors such as biosafety,materials,surgical safety,and manufacturing technique,currently the size of the implanted electrode array and the number of electrodes are extremely limited.So the artificial vision constructed by visual prosthesis has the problems of low visual acuity and small visual field.Therefore,how to extract useful information in the scene by effective image processing method under a small field and low-resolution artificial vision is an important problem to be solved in the field of visual prosthesis.Aiming at this problem,this paper proposes a global saliency detection method based on three visual features: color,color intensity difference and spatial distribution.This method can effectively extract and enhance salient objects in natural scenes.In order to verify the superiority of our method,we tested our method and other 10 state-of-the-art saliency methods on a public dataset,and the results show that our method has obvious superiority under several evaluation indicators which widely used in this field.In addition,this paper also initially builds an embedded image processing platform,and applies the saliency detection method proposed in this paper to the platform.We evaluate the performance and analyze the influencing factors of this algorithm under the embedded platform.It lays a foundation for the development of visual prosthesis image processing unit and the construction of image processing unit under simulated prosthetic vision in the future work.