Image Denoising with Circulant Representation and Haar Transform
Analysis
This paper introduces a computationally efficient image denoising algorithm, Haar-tSVD, that leverages the connection between PCA and the Haar transform within a circulant representation. The method's strength lies in its simplicity, parallelizability, and ability to balance speed and performance without requiring local basis learning. The adaptive noise estimation and integration with deep neural networks further enhance its robustness and effectiveness, especially under severe noise conditions. The public availability of the code is a significant advantage.
Key Takeaways
- •Proposes a computationally efficient image denoising algorithm (Haar-tSVD).
- •Establishes a theoretical connection between PCA and Haar transform under circulant representation.
- •Employs a unified tensor singular value decomposition (t-SVD) projection with Haar transform.
- •Offers a balance between denoising speed and performance without local basis learning.
- •Includes an adaptive noise estimation scheme for improved robustness.
- •Integrates deep neural networks to enhance performance under severe noise.
- •Provides publicly available code.
“The proposed method, termed Haar-tSVD, exploits a unified tensor singular value decomposition (t-SVD) projection combined with Haar transform to efficiently capture global and local patch correlations.”