Adaptive Filters to remove Blurring Effects over Time for Underwater Surveillance
Konstantinos Vougioukas · 2012
Exploring the ocean bottom has always been an area of great scientific and environmental concern. However, study of the underwater environment up until recently was very difficult due to the extreme conditions. With the advances in underwater photographic equipment surveillance of the sea bed is now easily realizable. The quality, however, of underwater images is still worse than that of images shot in the air and images usually appear hazy. This thesis deals with the problem of underwater surveillance of a scene. The quality of the recording obtained by the camera deteriorates over time due to problems like dirt/water on the lens and the glass protecting the camera, which is why the camera must be cleaned regularly. The dirt on the lens as well as floating particles create a blur and noise in the frames of the video. This projects’ main goal is to remove the blur effects from the underwater videos. As a secondary goal we wish to develop a method that uses the temporal information of the video as well as the knowledge of when the camera was cleaned. The method proposed in this study solves the problem in two stages. It first removes any noise that is present in the recordings and then deals with the blur effects. For the denoising stage a variation of the BM3D algorithm [8] was developed. Several different approaches were implemented for the deblurring problem based on the multiframe blind deconvolution method described in [1]. Evaluation of the algorithms was held for both artificial and real degradation of the frames.