Iterative Restoration of Noisy Elastically Distorted Periodic Images
M. Unser, B.L. Trus, M. Eden
Signal Processing, vol. 17, no. 3, pp. 191-200, July 1989.
The authors present several improvements to an approach that corrects for spatial distortion in quasi-periodic structures in order to achieve noise reduction by averaging. The warping function is represented by quasi-Hermite two-dimensional polynomials, a representation that allows great flexibility in the choice of fiduciary points. The estimation of the warping function is refined iteratively. At each iteration, the polynomial coefficients are evaluated from the current position of the reference points; the latter are then relocated by a cross-correlation technique. This procedure is intended to maximize a global signal-to-noise ratio criterion. Experiments with electron micrographs of thin sections of muscle fibers indicate a significant improvement in signal quality when compared with a previous approach. The new method is also shown to be insensitive to the initial position of the reference points.
@ARTICLE(http://bigwww.epfl.ch/publications/unser8903.html, AUTHOR="Unser, M. and Trus, B.L. and Eden, M.", TITLE="Iterative Restoration of Noisy Elastically Distorted Periodic Images", JOURNAL="Signal Processing", YEAR="1989", volume="17", number="3", pages="191--200", month="July", note="")