Switching bilateral filter with a texture/noise detector for universal noise removal

Area of Science:

• Image Processing
• Computer Vision
• Signal Processing

Background:

• Image noise significantly degrades visual quality and hinders subsequent analysis.
• Existing filters often struggle with mixed noise types (e.g., Gaussian and impulse).
• Efficient and accurate noise removal is crucial for various applications.

Purpose of the Study:

• To propose a universal noise removal method capable of handling diverse noise types.
• To develop a novel switching bilateral filter (SBF) integrated with a texture and noise detector.
• To achieve high-fidelity image restoration with reduced computational cost.

Main Methods:

• A two-stage approach: noise detection followed by filtering.
• Utilized the sorted quadrant median vector (SQMV) scheme for edge and texture information extraction.
• Classified pixels into impulse noise, Gaussian noise, or noise-free categories.
• Implemented a switching bilateral filter that adapts its filtering mode based on noise classification.

Main Results:

• The noise detector demonstrated high detection and classification rates for salt-and-pepper, uniform impulse, and mixed impulse noise.
• The SBF effectively removed both Gaussian and impulse noise, including mixed types.
• Achieved high peak signal-to-noise ratio (PSNR) and superior image quality compared to other filters.
• Demonstrated significantly lower computational complexity than existing mixed noise filters.

Conclusions:

• The proposed SBF with its integrated noise detector offers a robust solution for universal noise removal.
• The method excels in preserving image details while effectively suppressing various noise types.
• SBF provides a computationally efficient and high-performance alternative for image denoising.