Quaternion-based deep image prior with regularization by denoising for color image restoration

作     者：Zhang, Qinghua He, Liangtian Gao, Shaobing Deng, Liang-Jian Liu, Jun 

作者机构：Anhui Univ Anhui Univ Ctr Appl Math Sch Math Sci Hefei 230601 Peoples R China Sichuan Univ Coll Comp Sci Chengdu 610065 Peoples R China Univ Elect Sci & Technol China Sch Math Sci Chengdu 611731 Peoples R China Northeast Normal Univ Sch Math & Stat Key Lab Appl Stat MOE Changchun 130024 Peoples R China 

出 版 物：《SIGNAL PROCESSING》 (Signal Process)

年 卷 期：2025年第231卷

核心收录：

学科分类：0808[工学-电气工程] 08[工学] 

基　　金：Province, PR China [2024AH050064] Excel-lent University Research and Innovation Teams in Anhui Province, PR China [2024AH010002] Sichuan-Chongqing Science and Technology Innovation Cooperation Plan, PR China [CSTB2024TIAD-CYKJCXX0029] 

主　　题：Color image restoration Quaternion representation Deep image prior Regularization by denoising Quaternion ADMM 

摘      要：Deep image prior (DIP) has demonstrated remarkable efficacy in addressing various imaging inverse problems by capitalizing on the inherent biases of deep convolutional architectures to implicitly regularize the solutions. However, its application to color images has been hampered by the conventional DIP method s treatment of color channels in isolation, ignoring their important inter-channel correlations. To mitigate this limitation, we extend the DIP framework from the real domain to the quaternion domain, introducing a novel quaternionbased deep image prior (QDIP) model specifically tailored for color image restoration. Moreover, to enhance the recovery performance of QDIP and alleviate its susceptibility to the unfavorable overfitting issue, we propose incorporating the concept of regularization by denoising (RED). This approach leverages existing denoisers to regularize inverse problems and integrates the RED scheme into our QDIP model. Extensive experiments on color image denoising, deblurring, and super-resolution demonstrate that the proposed QDIP and QDIP-RED algorithms perform competitively with many state-of-the-art alternatives, both in quantitative and qualitative assessments. The code and data are available at the website: https://***/qiuxuanzhizi/QDIP-RED.