This paper addresses problems appearing in restoration algorithms based on utilizing both Tikhonov and bilateral total variation (BTV) regularization. The former regularization assumes that prior information has Gaussian distribution which indeed fails at edges, while the later regularization highly depends on the selected bilateral filter's parameters. To overcome these problems, we propose a locally adaptive regularization. In the proposed algorithm, we use general directional regularization functions with adaptive weights. The adaptive weights are estimated from local patches based on the property of the partially restored image. Unlike Tikhonov regularization, it can avoid smoothness across edges by using adaptive weights. In addition, unlike BTV regularization, the proposed regularization function doesn't depend on parameters' selection. The convexity conditions as well as the convergence conditions are derived for the proposed algorithm.
The copyright of the original papers published on this site belongs to IEICE. Unauthorized use of the original or translated papers is prohibited. See IEICE Provisions on Copyright for details.
Copy
Osama AHMED OMER, Toshihisa TANAKA, "Image Restoration Based on Adaptive Directional Regularization" in IEICE TRANSACTIONS on Fundamentals,
vol. E92-A, no. 12, pp. 3344-3354, December 2009, doi: 10.1587/transfun.E92.A.3344.
Abstract: This paper addresses problems appearing in restoration algorithms based on utilizing both Tikhonov and bilateral total variation (BTV) regularization. The former regularization assumes that prior information has Gaussian distribution which indeed fails at edges, while the later regularization highly depends on the selected bilateral filter's parameters. To overcome these problems, we propose a locally adaptive regularization. In the proposed algorithm, we use general directional regularization functions with adaptive weights. The adaptive weights are estimated from local patches based on the property of the partially restored image. Unlike Tikhonov regularization, it can avoid smoothness across edges by using adaptive weights. In addition, unlike BTV regularization, the proposed regularization function doesn't depend on parameters' selection. The convexity conditions as well as the convergence conditions are derived for the proposed algorithm.
URL: https://global.ieice.org/en_transactions/fundamentals/10.1587/transfun.E92.A.3344/_p
Copy
@ARTICLE{e92-a_12_3344,
author={Osama AHMED OMER, Toshihisa TANAKA, },
journal={IEICE TRANSACTIONS on Fundamentals},
title={Image Restoration Based on Adaptive Directional Regularization},
year={2009},
volume={E92-A},
number={12},
pages={3344-3354},
abstract={This paper addresses problems appearing in restoration algorithms based on utilizing both Tikhonov and bilateral total variation (BTV) regularization. The former regularization assumes that prior information has Gaussian distribution which indeed fails at edges, while the later regularization highly depends on the selected bilateral filter's parameters. To overcome these problems, we propose a locally adaptive regularization. In the proposed algorithm, we use general directional regularization functions with adaptive weights. The adaptive weights are estimated from local patches based on the property of the partially restored image. Unlike Tikhonov regularization, it can avoid smoothness across edges by using adaptive weights. In addition, unlike BTV regularization, the proposed regularization function doesn't depend on parameters' selection. The convexity conditions as well as the convergence conditions are derived for the proposed algorithm.},
keywords={},
doi={10.1587/transfun.E92.A.3344},
ISSN={1745-1337},
month={December},}
Copy
TY - JOUR
TI - Image Restoration Based on Adaptive Directional Regularization
T2 - IEICE TRANSACTIONS on Fundamentals
SP - 3344
EP - 3354
AU - Osama AHMED OMER
AU - Toshihisa TANAKA
PY - 2009
DO - 10.1587/transfun.E92.A.3344
JO - IEICE TRANSACTIONS on Fundamentals
SN - 1745-1337
VL - E92-A
IS - 12
JA - IEICE TRANSACTIONS on Fundamentals
Y1 - December 2009
AB - This paper addresses problems appearing in restoration algorithms based on utilizing both Tikhonov and bilateral total variation (BTV) regularization. The former regularization assumes that prior information has Gaussian distribution which indeed fails at edges, while the later regularization highly depends on the selected bilateral filter's parameters. To overcome these problems, we propose a locally adaptive regularization. In the proposed algorithm, we use general directional regularization functions with adaptive weights. The adaptive weights are estimated from local patches based on the property of the partially restored image. Unlike Tikhonov regularization, it can avoid smoothness across edges by using adaptive weights. In addition, unlike BTV regularization, the proposed regularization function doesn't depend on parameters' selection. The convexity conditions as well as the convergence conditions are derived for the proposed algorithm.
ER -