Using uncorrupted neighborhoods of the pixels for impulsive noise suppression with ANFIS

Civicioglu P.

IEEE TRANSACTIONS ON IMAGE PROCESSING, vol.16, no.3, pp.759-773, 2007 (SCI-Expanded) identifier identifier

  • Publication Type: Article / Article
  • Volume: 16 Issue: 3
  • Publication Date: 2007
  • Doi Number: 10.1109/tip.2007.891067
  • Journal Indexes: Science Citation Index Expanded (SCI-EXPANDED), Scopus
  • Page Numbers: pp.759-773
  • Keywords: adaptive network-based fuzzy inference system (ANFIS), extreme value distribution (EVD), impulsive noise (IN), k-nearest pixels, statistical noise detection, HIGHLY DISTORTED IMAGES, MEDIAN FILTER, CORRUPTED IMAGES, DIGITAL IMAGES, COLOR IMAGES, FUZZY FILTER, REMOVAL, RESTORATION, REDUCTION, DETECTOR
  • Erciyes University Affiliated: Yes


In this paper, a novel adaptive network-based fuzzy inference system (ANFIS)-based filter, ABF, is presented for the restoration of images corrupted by impulsive noise (IN). The ABF is performed in two steps. In the first step, impulse detection is realized by using statistical tools. In the second step, a nonlinear filtering scheme based on ANFIS is performed for only the corrupted pixels detected in the first step. To demonstrate the effectivity of ABF at the removal of high-level IN, extensive simulations were realized for ABF and nine different comparison filters. Empirical results indicate that the proposed filter achieves a better performance than the comparison filters in terms of noise suppression and detail preservation, even when the images are highly corrupted by IN.