Speckle Noise Removal Model Based on Diffusion Equation and Convolutional Neural Network.

Area of Science:

• Computer Vision
• Image Processing
• Machine Learning

Background:

• Convolutional neural networks (CNNs) offer good image denoising but lack robustness.
• Diffusion equation-based methods provide stability and theoretical guarantees for image denoising.
• Combining CNNs and diffusion equations can leverage the strengths of both approaches.

Purpose of the Study:

• To develop a robust speckle noise denoising model by integrating CNNs and diffusion equations.
• To address the hyperparameter dependency on noise variance in CNN-based denoising.
• To improve the practical applicability and performance of image denoising techniques.

Main Methods:

• A neural network model incorporating residual and structure learning was developed based on speckle noise mathematical models and image decomposition.
• A nonlinear diffusion equation-based algorithm was proposed for accurate noise variance estimation.
• The final model synergistically combines the proposed neural network with the diffusion equation for speckle noise removal.

Main Results:

• The proposed noise variance estimation algorithm demonstrates high accuracy.
• Numerical simulations confirm the effectiveness of the combined model in speckle noise denoising.
• The method shows significant practical application value in image processing.

Conclusions:

• The hybrid CNN and diffusion equation model effectively denoises images with speckle noise.
• The noise variance estimation algorithm overcomes limitations of previous methods.
• This integrated approach offers a robust and accurate solution for speckle noise reduction.