Geman and Reynolds (ibid., vol.14, 1992) present an approach to linear image restoration which provides for recovery of horizontal and vertical gray-level discontinuities from blurred and noisy observations. We extend their model and parameter selection method to include diagonal discontinuities. A hazard of this modeling approach is identified and addressed. We also comment on the truncated Gibbs sampler suggested in the paper
Published in:
Pattern Analysis and Machine Intelligence, IEEE Transactions on
(Volume:18
,
Issue:
6
)
Date of Publication: Jun 1996