Pixel-Changes-Based RGB Channels for a Novel Underwater Image Dehazing Technique

The field of underwater imaging is plagued by the problem of haze, which is an obscuring effect caused by water molecules and suspended particles absorbing and scattering light rays before the rays reach the camera. This proposed technique exploits individual RGB color channels and analyzes small pixel deltas in order to dehaze underwater images and increase perceived visual accuracy. A GARCH model is employed to estimate transmission maps because it can accurately estimate human visual perception by exploiting the volatility of pixel changes within color channels that result from changes in pixels between enhanced image features. Background light is estimated using differential light emissions in different color wavelengths. The proposed system is evaluated based on entropy, UIQMnorm, and UCIQE, then compared with existing state-of-the-art methods. The results demonstrate that this technique is superior in both visual analysis and performance evaluation metrics. Future research should examine the effect of GARCH alone on the algorithm and also explore the use of other networks such as Hopfield neural networks to alternatively implement the algorithm.