Title
A non-linear filter for impulsive noise reduction
Abstract
A new generic recursive operator mainly conceived to eliminate impulsive (salt-and-pepper) noise is presented. This nonlinear operator is composed of two independent conditional rules which are applied in a two-step sequence. Each rule has one- and bi-dimensional versions. This operator intends to eliminate all noisy pixels, identifying them as the ones too different from their surrounding neighborhoods, and replacing their gray level with the one of the most conservative neighbor. In this way, no new values of gray levels appear in the resulting image. This operator can be decomposed at different levels, resulting in a set of one dimensional operators. We have compared the performance of this new operator with classical operators using natural and artificial images and noise sources. We have found that it performs better in terms of image difference. Also, it is simple to implement, fast, and scalable to be adapted to particular applications.
Publication Date
12-1-1998
Publication Title
Proceedings of the Joint Conference on Information Sciences
Volume
4
Number of Pages
429-432
Document Type
Article; Proceedings Paper
Personal Identifier
scopus
Copyright Status
Unknown
Socpus ID
1642365799 (Scopus)
Source API URL
https://api.elsevier.com/content/abstract/scopus_id/1642365799
STARS Citation
Windyga, Piotr S., "A non-linear filter for impulsive noise reduction" (1998). Scopus Export 1990s. 3628.
https://stars.library.ucf.edu/scopus1990/3628