6 September 2017 An underwater turbulence degraded image restoration algorithm
Author Affiliations +
Abstract
Underwater turbulence occurs due to random fluctuations of temperature and salinity in the water. These fluctuations are responsible for variations in water density, refractive index and attenuation. These impose random geometric distortions, spatio-temporal varying blur, limited range visibility and limited contrast on the acquired images. There are some restoration techniques developed to address this problem, such as image registration based, lucky region based and centroid-based image restoration algorithms. Although these methods demonstrate better results in terms of removing turbulence, they require computationally intensive image registration, higher CPU load and memory allocations. Thus, in this paper, a simple patch based dictionary learning algorithm is proposed to restore the image by alleviating the costly image registration step. Dictionary learning is a machine learning technique which builds a dictionary of non-zero atoms derived from the sparse representation of an image or signal. The image is divided into several patches and the sharp patches are detected from them. Next, dictionary learning is performed on these patches to estimate the restored image. Finally, an image deconvolution algorithm is employed on the estimated restored image to remove noise that still exists.
Conference Presentation
© (2017) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Md. Hasan Furhad, Murat Tahtali, Andrew Lambert, "An underwater turbulence degraded image restoration algorithm", Proc. SPIE 10410, Unconventional and Indirect Imaging, Image Reconstruction, and Wavefront Sensing 2017, 104100R (6 September 2017); doi: 10.1117/12.2277179; https://doi.org/10.1117/12.2277179
PROCEEDINGS
6 PAGES + PRESENTATION

SHARE
Back to Top