FPGA IMPLEMENTATION OF ADAPTIVE HAZE REMOVAL OF AN IMAGE

Abstract

The presence of haze in images significantly degrades their quality by reducing detailing, contrast, and adding noise. Effective haze removal, or dehazing, is crucial for various computer vision applications such as autonomous navigation, surveillance, and remote sensing. This thesis presents a novel adaptive haze removal technique utilizing the YIQ colour space, which significantly aids in retaining the actual colour of the image. Inspired by the existing Dark Channel Prior (DCP) algorithm that operates in the RGB colour space, our approach adapts this methodology to the YIQ colour space. Additionally, we reference various algorithms that have successfully improved the quality of dehazed images to develop our YIQ-based dehazing algorithm. Key challenges addressed include estimating global atmospheric light, local atmospheric light, and the transmission map. The process begins by converting the image from RGB to YIQ colour space. The luminance component (Y) is then processed to estimate global atmospheric light, local atmospheric light, and edges within the image. Using these parameters, the transmission map is deduced. I and Q components to estimate the colour components of the image. To further enhance the transmission map and improve haze estimation, a morphological process is applied.