Images acquired through underwater turbulent media make the image processing tasks in image restoration and object identification challenging. Turbulence in water is associated with random fluctuations of temperature and salinity. These fluctuations are responsible for changing the refractive index, for attenuating illumination, imposing geometric distortions and space-variant blur on images, thus making object identification more difficult. In this paper, we propose a patch-wise deconvolution procedure for removing the space-variant blur from images for restoration purpose prior to resolving the object identification issue. The deconvolution procedure is aided with an image alignment procedure for obtaining better results. Next, an image segmentation algorithm based on fuzzy clustering is considered for object identification. Computational experiments are conducted using a real-world dataset to demonstrate the efficiency of the proposed method.
|