Efficient restoration of space-variant blurs from physical optics by sectioning with modified Wiener filtering

Authors

    Authors

    T. P. Costello;W. B. Mikhael

    Comments

    Authors: contact us about adding a copy of your work at STARS@ucf.edu

    Abbreviated Journal Title

    Digit. Signal Prog.

    Keywords

    image restoration; space-variant point-spread function; physical optics; sectioning; Wiener filtering; IMAGE-RESTORATION; CONVEX PROJECTIONS; SYSTEMS; Engineering, Electrical & Electronic

    Abstract

    Digital images bluffed by space-variant point-spread functions of uncorrected physical optics may be efficiently restored using Wiener filtering on overlapping subimage frames. Frame size must be tailored to accommodate the displacement-dependent spreading and shifting of physical point-spread functions at large field angles to prevent circular convolution edge effects from corrupting the frame's central section. We define the section as the nonoverlapping subframe used to construct the composite full-image restoration. Otherwise, if the frame is too small, edge-effect errors may extend into the section, inducing artifacts in the composite restoration. Conversely, if the frame is too large, total restoration processing will be greater than necessary. By adjusting frame size with field displacement, we demonstrate the effective restoration of images blurred by a laboratory-grade spherical lens. Blurred images are simulated and then restored with a modified Wiener filter. Mean-square-error and restoration improvement are reported as a function of field angle and criteria are developed for frame and section size selection. (C) 2002 Elsevier Science (USA). All rights reserved.

    Journal Title

    Digital Signal Processing

    Volume

    18

    Issue/Number

    1

    Publication Date

    1-1-2003

    Document Type

    Article

    Language

    English

    First Page

    1

    Last Page

    22

    WOS Identifier

    WOS:000185827300001

    ISSN

    1051-2004

    Share

    COinS