Infrared object detection via patch‐tensor model and image denoising based on weighted truncated Schatten‐p norm minimization. Issue 6 (10th February 2023)

• Record Type: Journal Article
• Authors: Zhu, Yun; Gong, Chengjian; Liu, Shuwen; Yu, Zhiyue; Shao, Hanzeng; Yu, Gaohang
• Is Part Of: IET image processing. Volume 17:Issue 6(2023)
• Page Start: 1762
• Page End: 1774
• Publication Date: 2023-02-10
• Subjects: alternating direction multiplier -- infrared object detection -- low‐rank -- non‐local denoising -- Schatten‐p norm; Image processing -- Periodicals; 621.36705
• DOI: 10.1049/ipr2.12753
• Languages: English
• ISSNs: 1751-9659

Abstract

Abstract: The nuclear norm minimization (NNM) is a special non‐convex rank minimization convex relaxation scheme for image denoising and object detection that requires denoising and background subtraction. Considering excessive shrinkage of rank components and equal treatment of different rank components, NNM is extended to the weighted Schatten‐ p norm minimization (WSNM) with weights assigned to different singular values. In this paper, a multi‐channel weighted truncated WSNM model based on the WSNM optimization framework is proposed for RGB colour image denoising. On the basis of different noise intensities and non‐local self‐similar patches of each channel of the colour image itself, the proposed model is improved significantly by the optimization methods of superposition and truncation. Meanwhile, it can be generalized to the tensor space and employed to the infrared imaging target detection based on the spatial‐temporal tensor model for the first time. And the efficient alternating direction multiplier‐based algorithms are developed to solve the proposed model and the accuracy of the algorithm is effectively improved by choosing an adaptive threshold. Extensive experiments on real infrared data verified the proposed method state‐of‐the‐arts and effectiveness. Abstract : A multi‐channel Weighted Truncated Schatten‐p Minimization (MCWTSNM) model for RGB color image denoising is proposed based on the different noise intensities and non‐local self‐similar patches of each channel of the color image itself.Meanwhile, it is generalized for the first time to the tensor space and applied to the infrared imaging target detection based on the spatiotemporal tensor model.