Fuzzy anisotropic diffusion based on edge detection

Jialin Song, H. R. Tizhoosh

Research output: Contribution to journalArticlepeer-review

Abstract

A fuzzy anisotropic diffusion algorithm based on edge detection and noise estimation is proposed for image denoising and edge enhancement. The edginess and noisiness fuzzy membership values are calculated with the edge detector and noise deviation of center pixel from the neighboring average, respectively. The employed edge detector provides more accurate estimation of edges and is less sensitive to noise than the gradient operator in anisotropic diffusion. Taking noise into account ensures that the diffusion process works well regardless of the type of noise degradation, and effectively reduces the number of iterations. We demonstrate how the rather complicated edge detection and noise estimation can be put together through fuzzy inference and embedded into anisotropic diffusion to provide better control on the diffusion processing. Quantitative and qualitative evaluations demonstrate superior performance of the proposed fuzzy approach while processing images with additive and multiplicative noise.

ASJC Scopus subject areas

  • Statistics and Probability
  • Engineering(all)
  • Artificial Intelligence

Fingerprint

Dive into the research topics of 'Fuzzy anisotropic diffusion based on edge detection'. Together they form a unique fingerprint.

Cite this