A Decision-Based Modified Total Variation Diffusion Method for Impulse Noise Removal.

Area of Science:

• Image processing
• Computer vision
• Signal processing

Background:

• Traditional impulsive noise removal methods like median filtering and total variation L1 can cause excessive smoothing and feature blurring.
• These methods are often limited to images with low-density noise, failing to preserve visual quality in more challenging scenarios.

Purpose of the Study:

• To develop an advanced impulsive noise removal technique that overcomes the limitations of existing methods.
• To enhance image quality by preserving visual features during the denoising process, especially under high noise densities.

Main Methods:

• A novel method that categorizes pixels based on noise characteristics: corrupted, noise-free, and possibly corrupted.
• Application of modified total variation diffusion to corrupted pixels and weighted total variation diffusion to possibly corrupted pixels.
• Pixels are processed differently based on their assigned category, leaving noise-free pixels unchanged.

Main Results:

• The proposed method demonstrates robustness across various noise strengths and image types.
• Experimental results show superior performance in noise removal compared to traditional methods, evidenced by higher PSNR/SSIM values.
• Restored images exhibit significantly improved visual quality with preserved details.

Conclusions:

• The new pixel-categorization-based denoising method effectively removes impulsive noise while minimizing feature blurring.
• This technique offers a robust and versatile solution for image denoising, suitable for diverse applications and noise levels.