Title: Classification and feature extraction of binucleate cells using Mahalanobis distance and Gabor wavelet analysis
Author: Kahkashan Afrin
Address: Department of Electronics and Communication Engineering, Birla Institute of Technology, Mesra-835215, Jharkhand, India
Abstract: A hybrid methodology of feature extraction and classification is proposed in this paper to classify binucleate and non-binucleate or normal cells. The proposed methodology consists of a Gabor filter-based feature extraction using two types of Gabor filters, namely circular Gabor and Gabor wavelet. Feature matrix considering mean and variance are calculated in sets of 50 for each of the filters. Thereafter, dimensionality reduction is done using a binary particle swarm optimisation (BPSO) technique to screen out the redundant features. Finally, Mahalanobis distance (MD) is used to classify the images into respective classes using the reduced set of features. To show the efficacy and robustness of the proposed hybrid technique using Gabor wavelets, the classification accuracy is calculated and compared with circular Gabor.
Keywords: Gabor filters; circular Gabor; Gabor wavelet analysis; Gabor features; binary PSO; particle swarm optimisation; BPSO; Mahalanobis distance; binucleate cells; non-binucleate cells; cell classification; feature extraction.
Int. J. of Intelligent Engineering Informatics, 2014 Vol.2, No.4, pp.304 - 324
Date of acceptance: 24 Nov 2014
Available online: 23 Jan 2015