Adaptive multi-domain framework for noise classification and image denoising
Abstract
The real-world imaging systems is the most challenging task, as the images suffer from various types of noise that occur during acquisition or transmission. As the cause for each noise is different, their characteristics are also different. Hence, there arise the need for a method which can denoise the image irrespective of its characteristic and manage the image. To address this challenge, this study presents a three-stage image enhancement structure. First is the identification of the noise in an image, then based on the characteristic of the noise three different filters are applied parallelly to supress each characteristic. The output obtained from each filtered result is fused at pixel level through a reliability-based fusion strategy. The proposed work considers various concepts i.e., mathematical concepts, rank order guided by entropy, and the curvature constraints with variational optimization. In fusion method, decision mechanism based on the evidence is used as it provides a reliable image feature. The proposed technique is proved by considering 5 different noise models and the experimental outcomes confirmed at a noise level of sigma=25 with an average peak signal-to-noise ratio (PSNR) of 39.84 dB and a structural similarity index measure (SSIM) of 0.974. Also, the method consistently achieved improvements of approximately 3.2–3.7 dB.
Keywords
Curvature penalty; Dempster–Shafer fusion; Entropy-guided rank-order filtering; Image denoising; Monogenic signal analysis; Noise classification; Riemannian manifold diffusion
Full Text:
PDFDOI: https://doi.org/10.11591/eei.v15i6.14184
Refbacks
- There are currently no refbacks.

This work is licensed under a Creative Commons Attribution-ShareAlike 4.0 International License.
Bulletin of Electrical Engineering and Informatics (BEEI)
ISSN: 2089-3191
,
e-ISSN: 2302-9285
This journal is published by the
Institute of Advanced Engineering and Science (IAES)
.