Real-Time Visual Saliency Detection Using Gaussian Distribution

Abstract

Image visual saliency detection without prior knowledge of image details is fundamental for many computer vision tasks including object recognition, image retrieval, and image segmentation. In order to achieve more accurate and quick detection, this paper proposed a novel global contrast method to generate full resolution saliency maps using Gaussian distribution model. Compared with existing methods, this developed algorithm could be implemented in real-time with a higher accuracy. After a reasonable estimation of the parameters in our method, comparison experiments were conducted with five typical algorithms, experimental results demonstrate our approach is faster than the current real time approaches and accurate in maintaining high quality.