Quality Assessment of Sonar Images

Abstract

AbstractIn many underwater applications such as biometric tracking and underwater rescue, it is necessary to analyze sonar images that contain a large amount of important information like submarine geomorphology, marine organisms, and wreck remains. Traditional optical imaging and analysis methods are inapplicable due to the limited penetration depth of visible light, so sonar imaging techniques play important roles in marine applications. Sonar image signals usually need to be transmitted via the underwater acoustic channel (UAC) for further analysis. Nonetheless, the sonar images presented to users by the terminal are subject to typical underwater distortion because of the poor UAC condition. For the QA of sonar images, this chapter first shows the sonar image database and the full-reference QA methods on account of local entropy, and statistical and structural information to capture the underwater distortion in sonar images. Second, it introduces task- and perception-oriented reduced-reference QA methods on account of the human visual system to evaluate the sonar images with poor quality in complex environments. Third, to address the issue of inaccessibility of reference sonar images in dynamic underwater environments, it presents the no-reference QA method based on contour degradation measurement. In the end, the QA methods mentioned above are validated on relevant databases, and the necessity of establishing efficient sonar image QA methods is indicated.