Underwater Image Enhancement by Dark Channel Dehazing and Lightness-Domain Rayleigh CLAHE

Abstract

Water degrades an optical image in two ways at once. Absorption depends on wavelength and produces a colour cast , scattering by suspended particles adds a veil that flattens contrast. The tonal range therefore stays narrow and the texture appears soft. In this paper presents an image domain method that treats each defect in a separate stageand the colour restoration repairs the imbalance between the three bands first. A dark channel estimate then removes the veil at moderate strength so dark regions do not collapse towards black, the image moves to the HLS space where contrast limited adaptive histogram equalization with a Rayleigh target acts on the lightness channel alone. Hue and saturation pass through untouched and the restored balance survives the contrast stage, unsharp masking with a threshold closes the pipeline and recovers detail lost to forward scattering and the method needs no training data and no learned weights, nine parameters stay fixed across every image testing on the UIEB benchmark against four published methods shows a clear gain in structure, a verage gradient reaches 14.536 while the closest competitor reaches 7.080. Mean standard deviation reaches 40.466 against 31.453. Entropy and colourfulness also rank first once the dataset is averaged.