Tchebichef and Adaptive Steerable Based Total Variation Model for Image Denoising.

Area of Science:

• Image processing and computer vision
• Signal processing

Background:

• Image edges are crucial for human perception and visualization.
• Existing denoising methods often overlook edge orientation, leading to suboptimal results.
• Exploiting image sparsity in transform domains like DCT and wavelet improves denoising.

Purpose of the Study:

• To develop an advanced image denoising technique that effectively preserves and enhances structural information, particularly edges.
• To address the limitations of current methods by incorporating edge geometrical orientation into the denoising process.

Main Methods:

• Introduction of an adaptive steerable total variation regularizer (ASTV) based on geometric moments to denoise edges according to their orientation.
• Leveraging image sparsity in the orthogonal Tchebichef moment domain.
• Proposing a novel sparse regularizer combining Tchebichef moments and ASTV regularizers.
• Optimizing the denoising framework using a split Bregman-based multivariable minimization technique.

Main Results:

• The proposed ASTV regularizer effectively denoises edges by considering their geometrical orientation.
• The combined sparse regularizer utilizing Tchebichef moments and ASTV demonstrates improved denoising performance.
• Experimental results show the proposed method is competitive with existing techniques.

Conclusions:

• The novel denoising framework significantly boosts performance by incorporating edge orientation and image sparsity.
• The method achieves superior objective and subjective image quality compared to existing approaches.
• This work offers a more effective solution for preserving structural details during image denoising.