Data-Driven Tight Frame Learning Scheme Based on Local and Non-local Sparsity with Application to Image Recovery

Chen, Dai-Qiang

doi:10.1007/s10915-016-0205-x

Cited by 3 publications

(3 citation statements)

References 34 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Here we investigate the results with different values of I. Specifically, we fix M = 8 and set I ∈ [1,2,4,6,8,12] for the HSDA module. The special case of s = 1 denotes the full attention.…”

Section: Ablation Studymentioning

confidence: 99%

“…It cannot represent the more complex patterns in the texture images. In order to properly express the textures and tiny details of images, many approaches of learning adaptive filters from the image, such as dictionary learning and non-local techniques [1,2,7,8], have been widely investigated. Although such methods have shown relatively promising performance, they still suffer from several drawbacks: (1) they can only extract the shallow features which may limit further performance improvements; (2) the performance is highly sensitive to the noise disturbance.…”

Section: Introductionmentioning

confidence: 99%

See 1 more Smart Citation

Hybrid Sparse and Dense Attentions of Similar Regions for Image Denoising

Chen

2024

Preprint

View full text Add to dashboard Cite

Self-attention based on dot-product has achieved great success in the field of computer vision. Its effectiveness is owed to the large capacity of capturing long-range dependencies in feature maps. However, its quadratic computational complexity with respect to the image size hinders the further application of the self-attention modules. Therefore, a variety of strategies, which limit the regions of the computation of dot-product, have been proposed to reduce the computational amount. Through analyzing the advantages and disadvantages of these methods, we introduce a hybrid sparse and dense attention module (HSDA) which adopts the dense dot-product attention in most similar regions and the sparse attention in other regions. Numerical experiments on image denoising demonstrate that the designed HSDA module has the advantage of both sparse and local dense attentions, and can obtain similar PSNRs to full attention at lower computational amount. The corresponding network constructed by the HSDA modules can product favorable results compared to many state-of-the-art methods.

show abstract

Section: Ablation Studymentioning

confidence: 99%

Section: Introductionmentioning

confidence: 99%

Hybrid Sparse and Dense Attentions of Similar Regions for Image Denoising

Chen

2024

Preprint

View full text Add to dashboard Cite

show abstract

“…Bao et al showed the sub-sequence convergence of the iterative algorithm in [3], and gave a new globally convergent algorithm. Soon later, this data-driven tight frame model has been improved and applied to various problems [13,22,27,32,33,37,42,44]. However, as far as we know, these data-driven models mainly use L 2 fidelity, which is not suitable for the basic and significant impulsive noise removal problems.…”

Section: Introductionmentioning

confidence: 99%

Data-Driven Tight Frame Construction for Impulsive Noise Removal

Wu¹

2022

JCM

View full text Add to dashboard Cite

Wavelet frame-based image restoration using sparsity, nonlocal, and support prior of frame coefficients

Wang

Xiang

2017

Vis Comput

View full text Add to dashboard Cite

The wavelet frame systems have been widely investigated and applied for image restoration and many other image processing problems over the past decades, attributing to their good capability of sparsely approximating piece-wise smooth functions such as images. Most wavelet frame based models exploit the l 1 norm of frame coefficients for a sparsity constraint in the past. The authors in [50,17] proposed an l 0 minimization model, where the l 0 norm of wavelet frame coefficients is penalized instead, and have demonstrated that significant improvements can be achieved compared to the commonly used l 1 minimization model. Very recently, the authors in [11] proposed l 0 -l 2 minimization model, where the nonlocal prior of frame coefficients is incorporated. This model proved to outperform the single l 0 minimization based model in terms of better recovered image quality. In this paper, we propose a truncated l 0 -l 2 minimization model which combines sparsity, nonlocal and support prior of the frame coefficients. The extensive experiments have shown that the recovery results from the proposed regularization method performs better than existing state-of-the-art wavelet frame based methods, in terms of edge enhancement and texture preserving performance.

show abstract

Data-Driven Tight Frame Learning Scheme Based on Local and Non-local Sparsity with Application to Image Recovery

Cited by 3 publications

References 34 publications

Hybrid Sparse and Dense Attentions of Similar Regions for Image Denoising

Hybrid Sparse and Dense Attentions of Similar Regions for Image Denoising

Data-Driven Tight Frame Construction for Impulsive Noise Removal

Wavelet frame-based image restoration using sparsity, nonlocal, and support prior of frame coefficients

Contact Info

Product

Resources

About