Underwater Image Quality Database Towards Fish Detection

Abstract

Marine biosphere monitoring is one of the crucial links to the understanding and protection of the marine environment. Moreover, underwater images play an important role in marine biosphere monitoring. Since the capture and transmission conditions are extremely poor in complicated underwater environments, images will suffer typical types of distortions significantly. Pre-evaluation of the image quality to facilitate subsequent processing becomes particularly important. Traditional Image Quality Assessment (IQA) methods are normally developed based on perceptual quality. Nevertheless, images are captured for understanding and analysis to achieve the purpose of intelligent monitoring. There are barriers between traditional IQA and IQA in marine biosphere monitoring. To address this issue, this paper focused on the fish detection task which is a vital task in marine biosphere monitoring. An Underwater Image quality database for Fish Detection (UIFD) is proposed based on the characteristics of the underwater environment. This database can be utilized as a benchmark to develop and evaluate underwater IQA. In addition, it can also be used to provide guidance for improving the performance of underwater fish detection algorithms.