Monocular depth estimation of underwater reef images using color and texture cues

Abstract

Estimating the bathymetry of shallow underwater regions aids in crafting efficient strategies for managing and conserving coral reef ecosystems. This study presents an alternative approach for estimating bathymetry of single underwater coral reef images using color and texture information. The Red-Dark Channel Prior and Fast Fourier Transform algorithms were employed to extract this information from the images, followed by calibration to account for the impact of depth on color and texture. Both methods successfully generated absolute depth estimates with a maximum discrepancy of one meter from the ground truth. The presence of shadows, fine-textured surfaces, and excessive sunlight glare limits the methods. While limitations exist, the results are encouraging and demonstrate the potential of this approach for bathymetry estimation.