Deep Learning Based Switching Filter for Impulsive Noise Removal in Color Images

Area of Science:

• Computer Vision
• Image Processing
• Deep Learning

Background:

• Noise reduction is crucial for computer vision tasks like object detection.
• Deep learning significantly enhances image denoising performance.
• Existing deep learning methods primarily focus on Gaussian noise removal.

Purpose of the Study:

• To develop a deep learning-based technique for impulsive noise removal.
• To improve the performance of image denoising for impulsive noise.
• To address the limitations of current methods in handling non-Gaussian noise.

Main Methods:

• A switching filtering technique utilizing deep learning is proposed.
• A deep neural network detects distorted pixels.
• Detected pixels are restored using a fast adaptive mean filter.

Main Results:

• The proposed deep learning approach effectively removes impulsive noise.
• Experimental results demonstrate superior performance compared to state-of-the-art filters.
• The method shows significant improvements in color digital image denoising.

Conclusions:

• The developed deep learning switching filter is highly effective for impulsive noise removal.
• This technique offers a significant advancement over existing methods for impulsive noise suppression.
• The approach holds promise for enhancing computer vision system robustness.