Optimal bilateral filter and Convolutional Neural Network based denoising method of medical image measurements

•An innovative bio-inspired optimization based filtering system is considered.•Dragonfly and Modified Firefly algorithm is used to select the optimal parameters.•The Convolutional based Neural Network (CNN) is used to classify the denoised image as normal or abnormal. Image denoising has been foremo...

Full description

Saved in:
Bibliographic Details
Published in:Measurement : journal of the International Measurement Confederation 2019-09, Vol.143, p.125-135
Main Authors: Elhoseny, Mohamed, Shankar, K.
Format: Article
Language:English
Subjects:
Algorithms
Artificial neural networks
Bilateral filter
Classification
Classifiers
Denoising
Fitness
Image classification
Measurements
Medical image processing
Medical imaging
Neural networks
Noise reduction
Optimization
Parameter modification
Signal quality
Signal to noise ratio
Visual signals
Citations: Items that this one cites
Items that cite this one
Online Access:Get full text
Tags: Add Tag
No Tags, Be the first to tag this record!
Description
Summary:•An innovative bio-inspired optimization based filtering system is considered.•Dragonfly and Modified Firefly algorithm is used to select the optimal parameters.•The Convolutional based Neural Network (CNN) is used to classify the denoised image as normal or abnormal. Image denoising has been foremost concern in the field of medical imaging (MI). For image denoising, the most challenging is to protect the data bearing structures such as edges and surfaces to get good visual quality while enhancing Peak Signal to Noise Ratio (PSNR). In this work, an inventive bio-inspired optimization based filtering system is considered for the MI denoising process, the filter named as Bilateral Filter (BF). The execution of the denoising process influences the decision of selecting the optimal parameters, i.e., Gaussian and spatial weights. Here, these parameters are chosen by utilizing swarm based optimization that is Dragonfly (DF) and Modified Firefly (MFF) algorithm. For this parameter selection, multi-objective fitness function (PSNR and vector root mean square error (VRMSE)) is utilized. Moreover, Convolutional based Neural Network (CNN) classifier is used to classify the denoised image as normal or abnormal, with better classification rate. From the experimental results the proposed model, PSNR of 47.52 dB and error rate of 1.23 is attained contrasting to the existing filters and some classifiers.
ISSN:0263-2241
1873-412X
DOI:10.1016/j.measurement.2019.04.072