Lightweight and Real-Time Infrared Image Processor Based on FPGA

Area of Science:

• Electrical Engineering
• Computer Engineering
• Image Processing

Background:

• Infrared imaging systems often suffer from sensor non-uniformity and defective pixels, degrading image quality.
• Real-time processing of infrared images requires efficient algorithms and hardware acceleration.
• Existing solutions may have high resource overhead or power consumption.

Purpose of the Study:

• To develop a lightweight and real-time infrared image processor using FPGA technology.
• To implement hardware-oriented algorithms for non-uniformity correction and blind pixel compensation.
• To optimize the processor for minimal resource utilization and low power consumption.

Main Methods:

• Implementation of a two-point correction algorithm for sensor non-uniformity calibration.
• Development of a blind pixel detection algorithm using first-level approximation.
• Application of a side-window-filtering method for blind pixel compensation with simultaneous convolution kernel computation.
• Integration of lightweight histogram equalization for enhanced visual observation.
• Hardware implementation on a Xilinx XC7A100T-2 FPGA.

Main Results:

• Achieved real-time non-uniformity correction for 640x480 resolution images.
• Effective compensation of blind pixels while preserving image details.
• Processor utilized 10,894 LUTs, 9367 FFs, 4 BRAMs, and 5 DSP48.
• Operates at 30 frames per second with 1800 mW power consumption under a 50 MHz clock.
• Maximum operating frequency reached 186 MHz.

Conclusions:

• The proposed FPGA-based infrared image processor offers a lightweight and efficient solution for real-time image processing.
• The implemented algorithms provide effective non-uniformity correction and blind pixel compensation with minimal resource overhead.
• The design demonstrates lower power consumption compared to existing similar works, making it suitable for various infrared imaging applications.