Kimiyoshi MIYATA*,**, Norimichi TSUMURA*,
Hideaki HANEISHI* and Yoichi MIYAKE*
* Graduate School of Science and Technology,
Chiba University
1-33, Yayoi-cho, Inage-ku, Chiba 263-8522,
JAPAN
** Information Technology R&D Center,
Mitsubishi Electric Corporation
5-1-1, Ofuna, Kamakura, Kanagawa 247-8501,
JAPAN
Journal of The Society of Photographic Science and Technology of Japan, Vol.63, No. 1, pp.18-27(2000)
Abstract
We propose a new Wiener filtering method
that can improve the total quality of images
corrupted by additive noise without degrading
the sharpness caused by the noise reduction
process. The Wiener filters are designed
so as to minimize the mean square error between
the original and restored images in RGB color
space. The Wiener filters are calculated
from the covariance matrices of the observed
images on the basis of the assumptions that
the original image and noise have no correlation,
and the noise covariance can be estimated
at system characterization stage or from
a uniform density area in the image. The
covariance matrices of the observed images
are estimated from the neighboring pixels
which are selected around the current pixel
with a color classification technique. Restored
images by computer simulations were evaluated
both objectively and subjectively. As a result,
we confirmed the proposed method is effective
to improve the quality of noisy images compared
with the conventional filters.
Full paper ( to appeare )
[PDF][PDF figures]( in Japanese )