Image Filtering to Increase the Efficiency in the Construction of Underwater Photomosaics

Abstract This paper describes an innovative methodology designed to filter out images with redundant information, initially selected to form photo-mosaics of the seabed. Collecting visual data of the sea bottom using underwater robots usually entails hundreds or even thousands of images if missions are performed in large areas and/or have long durations. All these images usually present multiple overlapping parts that need to be filtered out wisely in order to avoid useless repetitive information, questionable pixel aggregations, unreal colour blends, or the collapse of the computational resources dedicated to the mosaic building process. Experiments have been done with datasets grabbed from an Autonomous Underwater Vehicle while observing marine habitats of special ecological interest. Results show important reductions of visual data destined to the mosaics without significant loss of relevant information to be included in them, saving, drastically, time and data processing efforts.