High-Radix Taylor-Optimized Tone Mapping Processor for Adaptive 4K HDR Video at 30 FPS

Area of Science:

• Computer Vision
• Image Processing
• Hardware Acceleration

Background:

• High Dynamic Range (HDR) imaging offers superior visual fidelity but faces real-time processing challenges.
• Existing Low Dynamic Range (LDR) systems and complex tone mapping algorithms limit practical HDR applications.
• The demand for real-time HDR processing is growing in computer vision, photography, and medical imaging.

Purpose of the Study:

• To develop an adaptive HDR tone mapping processor with high computational efficiency and robust image quality.
• To enable consistent performance across diverse and fluctuating illumination conditions.
• To accelerate HDR processing for high-resolution, real-time applications.

Main Methods:

• An exposure-adaptive factor integrated into a bilateral filtering framework dynamically optimizes parameters.
• A high-radix Taylor expansion technique accelerates floating-point logarithmic and exponential operations.
• Implementation on a Xilinx XCVU9P FPGA for hardware acceleration.

Main Results:

• The processor achieves 4K video processing at 30 frames per second (FPS) and operates at 250 MHz.
• Demonstrates superior image fidelity with an average Tone Mapping Quality Index (TMQI) of 0.9314.
• Achieves 43% reduction in logic resources compared to state-of-the-art designs.

Conclusions:

• The proposed adaptive HDR tone mapping processor offers significant improvements in throughput and hardware efficiency.
• Enables real-time HDR processing for demanding high-resolution applications.
• Outperforms existing solutions in both performance and resource utilization.