A "salt and pepper" noise reduction scheme for digital images based on Support Vector Machines classification and...

Area of Science:

• Computer Vision
• Image Processing
• Machine Learning

Background:

• Digital images are often corrupted by impulse noise, commonly known as "salt and pepper" noise.
• Existing noise reduction algorithms have limitations in effectively handling varying noise levels.

Purpose of the Study:

• To develop and evaluate a new impulse noise removal technique for digital images.
• To leverage Support Vector Machines (SVM) for both identifying and reconstructing noisy pixels.

Main Methods:

• A novel technique utilizing Support Vector Machines (SVM) for impulse noise removal.
• Employing SVM classification to detect noisy pixels and SVM regression for pixel value reconstruction.
• Generating synthetic training data for SVM to ensure controlled quality and complexity.
• Comparing a modified median filter and a regression-based filter against state-of-the-art algorithms.

Main Results:

• The proposed SVM-based filters demonstrated superior performance compared to existing state-of-the-art algorithms for low and medium impulse noise ratios.
• The technique achieved comparable results to other methods on highly corrupted images.
• Both classification and regression approaches within the SVM framework proved effective for noise reduction.

Conclusions:

• The presented Support Vector Machine (SVM) based impulse noise removal technique is effective and efficient.
• This method offers a promising alternative for digital image denoising, particularly for "salt and pepper" noise.
• The approach shows potential for further development in advanced image restoration tasks.