A Second-Order Method for Removing Mixed Noise from Remote Sensing Images

Area of Science:

• Remote Sensing
• Image Processing
• Computer Vision

Background:

• Remote sensing images are crucial for various applications.
• Common noise types include Gaussian and salt-and-pepper noise.
• Existing denoising methods struggle with mixed noise, leading to detail loss.

Purpose of the Study:

• To propose a robust method for remote sensing image denoising.
• To address limitations of current algorithms in handling mixed noise.
• To improve the preservation of image edges and textures.

Main Methods:

• A two-stage denoising approach was developed.
• Stage 1: Modified DnCNN with dilated convolution and DropoutLayer for initial noise reduction.
• Stage 2: Adaptive median filtering with weighted nearest neighbor pixels for refinement.

Main Results:

• The proposed method effectively reduces mixed Gaussian and salt-and-pepper noise.
• Subjective and objective evaluations show superior performance over traditional methods.
• Denoised images exhibit enhanced clarity, natural details, and preserved edge/texture features.

Conclusions:

• The second-order denoising method significantly improves remote sensing image quality.
• It offers a more effective solution for mixed noise compared to existing techniques.
• The method successfully balances noise removal with detail preservation.