Research On Saliency Detection Methods Based On The Visual Sensitivities Of RGB Colors

Saliency detection is one of the important issues in image processing and computer vision. In this paper, the sensitivity differences of Human Eyes to RGB Colors are introduced into saliency detection methods, and an improved saliency detection method based on RGB sensitivities is discussed. The work has significance for the development of techniques in image processing and computer vision.The current saliency detection methods uniformly quantize RGB channels to achieve the purpose of reducing the amount of computation, no considering the sensitivity differences of human eyes to RGB colors. In this paper, we introduce the sensitivities of human eyes into saliency detection methods, and design the improved saliency detection methods based on RGB sensitivity. First, we design the sensitivity test experiments of human eyes to RGB colors, obtain the sensitivity empirical values of human eyes to RGB colors through a large number of experiments, and apply the sensitivity empirical values to image quantization. And then, the non-uniform quantization method of RGB channels is presented based on the test results of RGB sensitivity, and the computing methods of the sensitivity-weighted color distance and the sensitivity-weighted region distance are presented. Next, the sensitivity-weighted color distance and region distance are applied to the classical saliency detection methods based on global contrast and region contrast, and the improved saliency detection methods based on the sensitivity-weighted color distance and the sensitivity-weighted region distance are proposed. Finally, the proposed saliency detection methods are tested on the MSRA-1000 dataset. Experimental results illustrated that the proposed methods have improved the performance of saliency detection to a certain extent.The non-uniform quantization method for RGB channels and the computing methods of the sensitivity-weighted color distance and the sensitivity-weighted region distance can be applied to many applications in image processing and computer vision, which can reduce the computational complexity and improve the performance of saliency detection.