Image Saliency Detection Based on Background Prior and Multi-feature Fusion

Abstract

Image saliency region calculation is a very challenging computer vision problem. It is difficult to highlight the salient object uniformly with a single feature, integrating more effective prior knowledge is the feasible direction to improve the saliency region detection. In order to solve the problem of poor detection effect of image with complex background, this study proposes an image saliency detection algorithm based on background prior and multi feature fusion. Firstly, the color, texture, edge and other feature information of the original image are extracted, and the original image is divided into different size and non-overlapping image patch structure by synthesizing the color and spatial distribution of pixels. Then, taking the space distance between image patches and the space distance between image patches and the center of original image as the weight, the color, gray and background prior contrast of image patches are calculated, and the image saliency region is calculated by contrast fusion. Finally, different segmentation scales are used to construct the image patch structure, and the saliency is calculated respectively, and multi-scale fusion is carried out to enhance the saliency of the image. In order to evaluate the algorithm, it is tested on MSRA-1000, which is an international public data set. The experimental results show that the algorithm can better highlight the salient object, suppress the background noise, and get the saliency image which is more consistent with the visual perception.