Improving Visibility and Fidelity of Underwater Images Using an Adaptive Restoration Algorithm

When light is transmitted in water from a subject to an observer, it is scattered and absorbed by the unstable environment such as suspended particles and turbid water. Due to these phenomena, underwater images usually have poor quality including low contrast, blurring, darkness, and color diminishing. In this thesis, we propose a new underwater image restoration algorithm that consists of two major phases: visibility restoration and fidelity restoration. In the first phase, underwater images are observed similar to haze images because they have the same problems of low contrast and color shifting. This motivated us to use the haze removal technique, namely, dark channel prior, to dehaze underwater images. Subsequently, in the second phase, we equalize the color mean in each RGB (Red, Green, Blue) channel to balance the color. We then use the CLAHE (Contrast limited adaptive histogram equalization) method to enhance the local contrast of the image in the CIELAB color space. Finally, we perform histogram stretching on the S and I channels of the HSI (Hue, Saturation, Intensity) color space to make the image more natural. Preliminary results indicated that the proposed method effectively improved visibility and fidelity of underwater images.