Manifold regularization support vector machine-based image significance detection method

Abstract

The invention provides a manifold regulation support vector machine-based image significance detection method, and belongs to the technical field of computer vision. The invention provides a semi-supervised manifold regulation support vector machine-based image significance detection method, artificially marked truth values are not needed, only few training samples are needed and significant target detection is respectively carried out from the global and local angles of images. According to the method, few pseudo-marked samples are trained, so that the artificial workload and model training cost are decreased; manifold regularization matrixes are respectively constructed from the global and local angles, so that the highlighting and integrity of the detected significant targets are ensured; an optimization method is combined to furthermore optimize the significance detection results predicted by vector support machine models, so that the detection is corrector and the target areas are more highlighted and smooth; and in special images, the method also can preferably detect the images with oversized or undersized significant targets and the images with a plurality of targets.

Description

technical field [0001] The invention belongs to the technical field of computer vision and relates to the technical field of image information processing, in particular to an image saliency detection method based on a manifold regularization support vector machine model, which is suitable for the salient target detection of static images. Background technique [0002] With the development of computer technology and the popularization of digital electronic products, image resources are becoming more and more abundant, which satisfies people's massive collection and application of image information, but the problem of complicated and redundant information also follows. Inspired by the efficient visual information processing mechanism of the biological vision system, image saliency detection in the field of computer vision came into being. The salient things in the image scene are basically rich in the main information of the image. More and more researchers have begun to explo...

AI Technical Summary

Problems solved by technology

[0003] Although great research results have been achieved in the field of image saliency detection, there are still many problems that have not been resolved, such as the image contains multiple salient objects, the object scale is too large or too small, and the detection results of some algorithms are not very accurate.

In addition, some saliency detection algorithms adopt a supervised method, which requires a large number of artificially labeled real-value training samples to train the detection model, which is relatively expensive; some saliency detection algorithms only use local or global image perspectives Performing saliency detection, resulting in incomplete or insufficiently prominent detected saliency targets

Images