Enhancement of Underwater Video Mosaics for Post-Processing

Mosaics of seafloor created from still images or video acquired underwater have proved to be useful for construction of maps of forensic and archeological sites, species\u27 abundance estimates, habitat characterization, etc. Images taken by a camera mounted on a stable platform are registered (at first pair-wise and then globally) and assembled in a high resolution visual map of the surveyed area. While this map is usually sufficient for a human orientation and even quantitative measurements, it often contains artifacts that complicate an automatic post-processing (for example, extraction of shapes for organism counting, or segmentation for habitat characterization). The most prominent artifacts are inter-frame seams caused by inhomogeneous artificial illumination, and local feature misalignments due to parallax effects - result of an attempt to represent a 3D world on a 2D map. In this paper we propose two image processing techniques for mosaic quality enhancement - median mosaic-based illumination correction suppressing appearance of inter-frame seams, and micro warping decreasing influence of parallax effects