Scene-based dual domain non-uniformity correction algorithm for stripe and optics-caused fixed pattern noise removal

Area of Science:

• Infrared imaging
• Image processing
• Signal processing

Background:

• Non-uniformity is a persistent issue in infrared imaging, causing fixed pattern noise (FPN) that degrades image quality.
• Current scene-based non-uniformity correction (NUC) algorithms effectively address stripe FPN but fail with spatially continuous optical FPN.

Purpose of the Study:

• To propose a novel scene-based dual domain correction approach for simultaneously removing stripe and optical FPN in infrared images.
• To enhance the quality of infrared images by mitigating degradation caused by non-uniformity.

Main Methods:

• A dual domain correction strategy is employed, utilizing frequency domain gain correction and spatial domain offset correction.
• Stripe FPN is addressed by approximating the desired image with a guided filter and iteratively updating bias correction parameters.
• Optical FPN is removed by separating low-frequency noise using Fourier transform and updating gain correction parameters.
• Ghost artifacts are mitigated through an adaptive strategy adjusting learning rates and weights during parameter updates.

Main Results:

• The proposed dual domain approach effectively removes both stripe and optical FPN.
• Comprehensive evaluations show superior performance compared to existing methods on both real and simulated infrared videos.
• The method successfully mitigates ghost artifacts, further enhancing image correction fidelity.

Conclusions:

• The developed dual domain correction method offers a robust solution for infrared image non-uniformity.
• This approach significantly improves infrared image quality by addressing multiple sources of FPN.
• The findings have implications for enhancing the performance of various infrared imaging applications.