A Dual-Approach to Drone Image Enhancement: Software-Based Processing and FPGA Implementation

This work offers a thorough method for real-time dehazing of drone-captured images by different filtering techniques with post-processing improvements. Enhancing visibility and picture clarity in hazy situations is the main goal since it is essential for applications like environmental monitoring, navigation, and surveillance. In order to estimate the atmospheric light and transmission map, the suggested methodology makes use of the dark channel. A farrow filter and a guided filter are then applied to improve the transmission. Metrics including the PSNR, SSIM, Execution time, and MSE that are derived from Python-based image processing implementations are used to assess the effectiveness of the dehazing algorithms. The outcomes show notable gains in processing efficiency and image clarity. In order to accelerate the performance, this method was implemented on an FPGA based SoC. Power consumption and throughput of the FPGA - based solution were evaluated, demonstrating its effectiveness and appropriateness for real-time applications.