VBI-Accelerated FPGA Implementation of Autonomous Image Dehazing: Leveraging the Vertical Blanking Interval for Haze-Aware Local Image Blending

Real-time image dehazing is crucial for remote sensing systems, particularly in applications requiring immediate and reliable visual data. By restoring contrast and fidelity as images are captured, real-time dehazing enhances image quality on the fly. Existing dehazing algorithms often prioritize visual quality and color restoration but rely on computationally intensive methods, making them unsuitable for real-time processing. Moreover, these methods typically perform well under moderate to dense haze conditions but lack adaptability to varying haze levels, limiting their general applicability. To address these challenges, this paper presents an autonomous image dehazing method and its corresponding FPGA-based accelerator, which effectively balance image quality and computational efficiency for real-time processing. Autonomous dehazing is achieved by fusing the input image with its dehazed counterpart, where fusion weights are dynamically determined based on the local haziness degree. The FPGA accelerator performs computations with strict timing requirements during the vertical blanking interval, ensuring smooth and flicker-free processing of input data streams. Experimental results validate the effectiveness of the proposed method, and hardware implementation results demonstrate that the FPGA accelerator achieves a processing rate of 45.34 frames per second at DCI 4K resolution while maintaining efficient utilization of hardware resources.