Sonar Image Quality Assessment Based on Degradation Measurement

Abstract

In many (if not only) sonar applications, it is necessary to send sonar images (SIs) to a remote location for further analysis via underwater acoustic channel. Since bad channel condition only provides with limited bandwidth and unstable link, sonar images are very likely to suffer from different distortions via transmission. To this end, we develop a no-reference sonar image quality metric (NSIQM) for perceiving the quality of SIs. This metric evaluates image quality by measuring contour degradation degree between a test image and its associated filtered version. Features containing contour information are extracted from spatial and frequency domains. Besides, a bootstrap aggregating (bagging) based support vector regression (SVR) module is employed to build the relationship between features and subjective qualities of SIs. The results of experiments validate that the proposed metric is competitive with state-of-the-art reference-based quality metrics, and outperforms the latest reference-free competitors.