Binarization Algorithm Based on Side Window Multidimensional Convolution Classification.

Area of Science:

• Computer Vision
• Image Processing
• Space Technology

Background:

• In-orbit cameras suffer from image degradation due to uneven illumination and space radiation.
• This degradation results in inhomogeneous grayscale distribution, low contrast, and noisy images.
• Traditional binarization algorithms struggle with degraded images due to their reliance on statistical information and neglect of morphological properties.

Purpose of the Study:

• To propose an accurate image binarization algorithm for degraded space images.
• To address the limitations of traditional local binarization methods in handling space image artifacts.
• To develop a robust binarization technique insensitive to noise.

Main Methods:

• A novel binarization algorithm based on morphological classification is presented.
• An eight-dimensional side window filtering (SWF) kernel is employed to capture pixel morphological properties.
• Local thresholds are calculated based on the difference between positive and negative pixel types within a local window, filtered by positive pixels for morphological smoothness.

Main Results:

• The proposed algorithm achieves good binarization results on various degraded images.
• Quantitative evaluation using FM, PSNR, and DRD metrics shows superior performance compared to three classical techniques.
• The algorithm demonstrates high accuracy, robustness, and insensitivity to noise.

Conclusions:

• The morphological classification binarization algorithm effectively handles degraded space images.
• The SWF kernel-based approach provides a robust solution for accurate image binarization in challenging space environments.
• The method offers significant improvements over existing techniques for in-orbit image processing.