Image significance detection method based on region label fusion

Abstract

According to the image significance detection method based on region label fusion, a super-pixel segmentation algorithm is used for preprocessing an image, and the image is segmented into a pluralityof image region blocks; Using a Gaussian kernel function to obtain region similarity, using the region similarity to perform spectral clustering of the superpixel region, obtaining a label set of image segmentation, and storing boundary information of the image according to the label set; Obtaining salient features of the image, and fusing the salient features under a conditional random field model to obtain a roughness saliency map; Utilizing the label set to propagate the boundary information, and comparing and fusing the boundary information and the roughness saliency map to obtain reconstruction of the roughness saliency map; A self-adaptive threshold segmentation mode is adopted to carry out binarization processing on the reconstructed roughness saliency map, a label indication vectoris utilized to label a saliency region into a unified label, isolated points in the saliency region are processed, and more effective saliency region detection is obtained.

Description

Technical field [0001] The invention relates to the technical field of image processing, in particular to an image saliency detection method based on regional label fusion. Background technique [0002] The human visual system can obtain regions of interest based on visual attention mechanisms for different scenes. Each image contains one or more salient targets, and saliency detection is to imitate the visual attention mechanism to obtain important information in the image to improve the efficiency and accuracy of image processing. Image saliency detection has a wide range of applications in image annotation and retrieval, target recognition, image automatic cropping, image compression and other fields, and it is one of the hotspots of computer vision research. [0003] Starting from the visual attention mechanism, saliency detection can be divided into bottom-up detection model and top-down detection model. The bottom-up model mainly uses the underlying features such as color, ...

AI Technical Summary

Problems solved by technology

[0007] The purpose of the embodiments of the present invention is to provide an image saliency detection method based on region label fusion, which solves the problems of unclear boundaries and non-dense internal regions of image salient regions, and combines various salient features, conditional random field theory, and superpixel segmentation Method, spectral clustering and adaptive threshold segmentation are organically fused through label information to achieve more effective salient region detection

Images