Image saliency target detection method based on low-rank matrix recovery

Abstract

The invention discloses an image saliency target detection method based on low-rank matrix recovery, and the method comprises the steps: extracting color features from an original image, and determining a feature matrix of the original image in combination with image superpixels; decomposing a low-rank matrix from the feature matrix; constructing a hierarchical index tree of the original image byutilizing image superpixels according to an index tree generation algorithm, and determining a hierarchical sparse norm of the original image in combination with high-level prior information; performing ternary decomposition on the low-rank matrix, and determining a structured low-rank matrix recovery model of the original image; and fusing the layered sparse norm of the original image and the structured low-rank matrix recovery model of the original image, and combining an alternating direction optimization algorithm to obtain a saliency map. The method for accelerating singular value decomposition by low-rank matrix ternary decomposition is introduced, and solves the problem of high calculation complexity caused by minimization of matrix trace norm. By constructing an index tree and combining high-level prior, the problem of unsupervised image saliency target detection under a complex background is solved.

Description

technical field [0001] The invention relates to the technical field of image detection, in particular to an image salient target detection method based on low-rank matrix restoration. Background technique [0002] Salient object detection is effective to accurately segment foreground objects from a single scene, or to achieve collaborative detection from multiple images. At present, it has been widely used in image segmentation, content-based image retrieval, image compression, image cutting, etc. Salient object detection methods are broadly classified into supervised and unsupervised categories according to whether they utilize labeled information or not. Supervised methods usually utilize deep learning models and large-scale training algorithms to achieve object detection. Unsupervised methods do not require large-scale image samples, and have better flexibility and lower computational complexity. [0003] Unsupervised salient object detection methods usually exploit hi...

AI Technical Summary

Problems solved by technology

However, these existing methods usually use simple matrix norms to induce sparse matrices, ignoring the structural information of image saliency objects, resulting in divergent or incomplete phenomena in the generated saliency maps.

In addition, these methods use the matrix kernel norm to constrain the low-rank matrix, resulting in the algorithm needing to perform singular value decomposition in each iteration, increasing the computational cost

Images