When image through water turbulence, captured images will appear severe geometrical distortion. Restoration of image with unknown geometrical distortion is a challenging problem. In this paper, we propose an iterative method to acquire a geometrically corrected high-quality image from an observed video sequence which contains serious geometrical distortion. Firstly, we use a blind deconvolution method to deblur the temporal mean image of the observed video sequence. Next, an image registration based on B-spline is employed to obtain a new video sequence with less geometrical distortion. After several iterative computations of deconvolution and registration, we carry out a robust principal component analysis to remove residual noise of latest video sequence. Finally, a single geometrically corrected image is acquired by a temporal mean operation on final video sequence. Experimental results demonstrate that our proposed method is capable to greatly remove geometrical distortion in video sequence.
When imaging through an undulating clear water surface, a complex series of refractions and reflections occur throughout the imaging path, and light rays are bent by unknown amounts. So the captured images usually contain severe geometric distortions. In this paper, an iterative robust registration algorithm is employed to remove the distortions in frames by registering each frame to a reference image. As the traditional image registration algorithm is impeded by the severely blur mean, we decide to reconstruct a high quality reference as the surrogate of the mean. We first select the image patches with higher quality from the distorted sequence to reconstruct a single image. In the patches selection process, the image quality of patches is evaluated from sharpness and geometric distortion. Then the blind deconvolution technique is employed to deblur the image, which will be used as a reference of the next registration process. Experiment shows that the proposed algorithm performs well in restoring the distorted underwater images and has less computational time than the state-of-the-art method.
Access to the requested content is limited to institutions that have purchased or subscribe to SPIE eBooks.
You are receiving this notice because your organization may not have SPIE eBooks access.*
*Shibboleth/Open Athens users─please
sign in
to access your institution's subscriptions.
To obtain this item, you may purchase the complete book in print or electronic format on
SPIE.org.
INSTITUTIONAL Select your institution to access the SPIE Digital Library.
PERSONAL Sign in with your SPIE account to access your personal subscriptions or to use specific features such as save to my library, sign up for alerts, save searches, etc.