Research on FPGA-based voltage-adaptive EBAPS image denoising algorithm

Résumé fourni par la source

Electron Bombarded Active Pixel Sensor (EBAPS), a high-performance hybrid low-light-level imaging device, exhibits spatially shifting and flickering white noise that degrades image quality and increases with cathode voltage. Traditional denoising algorithms struggle with this voltage-dependent noise complexity. This paper proposes a Voltage-Dependent Adaptive Denoising Algorithm (VDADA) based on bilateral filtering. We first analyze the statistical distribution of EBAPS noise across different electron gain voltages, establishing a quantitative noise-voltage mapping relationship. Crucially, VDADA dynamically adjusts key bilateral filter parameters – window size, spatial standard deviation, and range standard deviation – in real-time based on the sensed cathode voltage and this mapping. The core principle employs milder filtering at low voltages to preserve detail, while implementing stronger filtering at high voltages to enhance suppression of shot noise and gain process noise. Compared to fixed-parameter algorithms, VDADA achieves excellent and balanced denoising performance across varying voltage conditions. Experimental results demonstrate significant noise reduction while effectively preserving edges and details, improving overall image quality. The algorithm achieves an average PSNR improvement of 3.09 dB and an average SSIM of 0.9894. This framework provides an effective approach for optimizing EBAPS and other imaging sensors with similar voltage-noise coupling characteristics, particularly suited for advanced high-sensitivity, wide-dynamic-range imaging applications.

Ce résumé expose les affirmations des auteurs. BNTIC ne l’interprète pas comme une validation indépendante des résultats.