Devising a comprehensive synthetic underwater image dataset

The underwater environment is characterized by complex light interactions, including effects such as color loss, contrast loss, water distortion, backscatter, light attenuation, and color cast, which vary depending on water purity, depth, and other factors. While many datasets in the literature contain specific ground-truth images, image pairs, or limited analysis with metrics, there is a need for a comprehensive dataset that covers a wide range of underwater effects with varying severity levels. This paper introduces a dataset consisting of 100 ground-truth images and 15,000 synthetic underwater images. Given the complexity of underwater light variations, simulating these effects is challenging. This study approximates the underwater effects using implementable combinations of color cast, blurring, low-light, and contrast reduction. In addition to generating 15,100 images, the dataset includes a comprehensive analysis with 21 focus metrics, such as the average contrast measure operator and Brenner’s gradient-based metric, as well as 7 statistical measures, including mean intensity and skewness.