An optimized pulse coupled neural network image de-noising method for a field-programmable gate array based...

Area of Science:

• Computer Vision
• Image Processing
• Artificial Intelligence

Background:

• Polarization image quality is often degraded by noise during acquisition.
• Existing de-noising methods may not adequately address noise in polarization imaging.
• Field-Programmable Gate Arrays (FPGAs) offer potential for real-time image processing.

Purpose of the Study:

• To propose and implement an optimized de-noising method for polarization images.
• To enhance the quality of images acquired by FPGA-based polarization cameras.
• To improve noise suppression and image fidelity in polarization imaging.

Main Methods:

• An adaptive Pulse Coupled Neural Network (PCNN) was employed for de-noising.
• PCNN parameters were optimized using Gray Wolf Optimization (GWO).
• Bi-Dimensional Empirical Mode Decomposition (BEMD) was utilized to decompose and simplify noisy images.
• The integrated method was implemented on an FPGA-based polarization camera.

Main Results:

• The proposed method effectively attenuated various noise types in polarization images.
• Significant improvements in image quality were observed, validated by metrics like PSNR and SSIM.
• The FPGA implementation enabled synchronous acquisition of high-quality, de-noised polarization images.
• The optimized PCNN approach outperformed existing state-of-the-art de-noising algorithms.

Conclusions:

• The combined GWO-PCNN and BEMD method offers a robust solution for polarization image de-noising.
• FPGA implementation facilitates real-time, high-quality polarization image acquisition.
• This technique substantially enhances the utility of polarization imaging in various applications.