A Novel Approach based on Probability Color Distribution and Color Models to Detect Salient Object

A salient object is the most important object in the scene which attracts to the human visual system. Salient object detection has been attracting a lot of interest, and recently various heuristic computational models have been designed. This paper presents a novel approach consisting of an algorithm based on RGB and HSV color models, and probability color distribution for detecting an appropriate salient object. Firstly, the proposed algorithm removes the background pixels of the image by subtracting the average mean of RGB channels from all the channels. Thus, fewer contributing regions from the channels are removed; still, a few non-salient regions remain. Then, the mean of the RGB component is subtracted from each RGB image respectively to remove less prominent pixels of each component and a most probable salient region is obtained after combining. This image is converted into HSV color. Hue and saturation have a strong representation of the salient object. The mean of HSV is computed to remove the non-salient region. The small mean image of HSV is multiplied by the maximum mean image of RGB to obtain higher probability color regions representing the salient region. Then, the salient region is pipelined to probability color distribution computation for non-salient region removal and finalizing correct detection of the salient object.