Saliency Detection using Gaussian Distribution

The methodology used to detect prominent regions of an image is known as Salient Image Detection. Saliency detection has become a major area of research, due to its application in computer vision, autonomous robots, medical imagery and also in data transmission. In this paper, we have put forth a method which capitalizes on the fact that salient regions of an image rarely touch the image boundaries. Our method computes the absolute difference between the mean boundary pixels and all the image pixels, followed by application of Gaussian distribution. We have performed experiments, after which it was concluded that our approach has a better quantitative and qualitative output than other popular saliency detection algorithms.