Efficient joint noise removal and multi exposure fusion.

Area of Science:

• Digital Image Processing
• Computer Vision
• Computational Photography

Background:

• Multi-exposure fusion (MEF) combines images with varying exposure times to create a single, well-exposed image.
• Existing methods often denoise images individually before fusion, which can be inefficient and suboptimal.
• Noise reduction is crucial for high-quality image reconstruction from multiple exposures.

Purpose of the Study:

• To propose a novel method for joint multi-exposure fusion and noise removal.
• To develop an efficient algorithm by performing fusion and denoising in the Discrete Cosinus Transform (DCT) domain.
• To significantly improve upon existing state-of-the-art MEF and denoising techniques.

Main Methods:

• A new strategy for joint fusion and denoising is presented.
• The algorithm operates within the Discrete Cosinus Transform (DCT) domain for computational efficiency.
• Key techniques include spatio-temporal patch selection and collaborative 3D thresholding.

Main Results:

• The proposed method effectively fuses multiple exposures while simultaneously removing noise.
• Joint processing in the DCT domain leads to a highly efficient algorithm.
• Experimental results demonstrate significantly superior performance compared to existing methods.

Conclusions:

• The developed technique offers an efficient and effective solution for simultaneous multi-exposure fusion and denoising.
• Joint optimization in the DCT domain is a promising approach for advanced image processing tasks.
• The method achieves state-of-the-art results, outperforming current techniques in image quality and efficiency.