论文 | Shuangli Du, Yiguang Liu, Minghua Zhao, Zhenyu Xu, Jie Li, Zhenzhen You：A new image decomposition approach using pixel-wise analysis sparsity model

本文（A new image decomposition approach using pixel-wise analysis sparsity  model）原载Pattern Recognition，由四川大学刘怡光教授等科研人员创作，系四川大学智慧法治超前部署学科系列学术成果。后续会持续分享四川大学智慧法治超前部署学科系列学术成果，欢迎大家阅读。

Abstract:Decomposing an image into two ‘simpler’ layers has been widely  used in low-level vision tasks, such as image recovery and enhancement. It is an  ill-posed problem since the number of unknowns are larger than the input. In this  paper, a two-step strategy is introduced, including task-aware priors estimate and a  decomposition model. A pixel-wise analysis sparsity model is proposed to regularize  8the separation layers, which supposes the transformed image generated with analysis  operator is sparse. Unlike regularizing all pixels with one penalty weight, we try to  estimate each pixel’s sparsity level with task-aware priors and to achieve pixel-wise  sparse penalty. Additionally, one separation layer is regularized with both synthesis  sparsity model and pixel-wise analysis sparsity model to exploit their  complementary mechanisms. Unlike the analysis one utilizing image local features,  the synthesis one exploits an over-complete dictionary and non-local similarity cues  to provide flexible prior for regularizing the decomposition results. The proposed  model is solved by an alternating optimization algorithm. We evaluate it with two  applications, Retinex model and rain streaks removal. Extensive experiments on  multiple enhancement datasets, many synthetic and real rainy images demonstrate  that our method can remove imaging noise during Retinex decomposition, and can  produce high fidelity deraining results. It achieves competing performance in terms  of quantitative metrics and visual quality compared with the state-of-the-art  methods.

Shuangli Du, Yiguang Liu, Minghua Zhao, Zhenyu Xu, Jie Li, Zhenzhen  You. A new image decomposition approach using pixel-wise analysis sparsity  model, Pattern Recognition, Pattern Recognition, Vol.136, Article 109241, April  2023..（论文下载）