Underwater Image Restoration: Effect of Different Dictionaries

Ocean engineering has a strong need for clear and high quality underwater images. Capturing a clear scene underwater is not a trivial task since color cast and scattering caused by light attenuation and absorption are common. The poor quality hinders the automatic segmentation or analysis of the images. In this work, an image restoration based on compressive sensing is reported which tackles with blurring caused by light scattering and provides better structural details. Furthermore, the effects of different dictionaries on the quality of restoration is studied. The aim is to use a single degraded underwater image and improve the image quality without any prior knowledge about the scene such as depth, camera-scene distance or water quality.