An improved saliency detection method based on non-uniform quantification and channel-weighted color distance

摘要

We propose a non-uniform quantification method for RGB channels based on the visual sensitivities of human eyes to the red, green and blue colors, and give weights to RGB channels which are used to compute the channel-weighted color distances. Furthermore, we design a saliency detection method using the non-uniform quantification method and the channel-weighted color distances measurement. We first quantify the values in RGB channels in different step-lengths and determine the weights w (r) , w (g) and w (b) based on the visual sensitivities of human eyes to the three-primary colors. And then we convert the quantified image into that in Lab color space, which are segmented into some regions, and compute the channel-weighted color distance using the weights w (L) , w (a) and w (b) , which are computed from the weights w (r) , w (g) and w (b) and the transformation from RGB color space to Lab color space. The saliency of each region is computed using the weighted distances and spatial weights. The proposed non-uniform quantification method and the channel-weighted color distance measurement can be used for the applications based on color features in the field of image processing and computer vision. Experimental results show that our methods can improve the efficiency and performance of saliency detection to some extent.