Indirect immunofluorescence image classification using texture descriptors
Texture descriptors, Local binary patterns, Support vector machine, Ensemble, HEp-2 cells classification
In this work we propose an ensemble of texture descriptors for HEp-2 cell classification. Our system is based on a "œpyramidal application" of local binary patterns coupled with a method for handling nonuniform bins. This feature extraction approach is then combined with a support vector machine (SVM) classifier. We test our method on a recent contest dataset (the MIVIA HEp-2 images dataset) using different testing protocols. This dataset is very challenging since the images are characterized by high variability in illumination. Therefore, to obtain good results, it is essential to apply a preprocessing algorithm: we choose the histogram equalization. We found that the best results are obtained when the original intensity images are converted into grayscale images with ten discrete values. Since a training set is provided in the contest dataset, we use it for descriptor selection and for parameter settings. The system built by using the training data is then applied to the testing set. Experiments show that our method outperforms the winner of the recent contest at the 21st International Conference on Pattern Recognition 2012.
Nanni, Loris, Michelangelo Paci, and Sheryl Brahnam. Indirect immunofluorescence image classification using texture descriptors." Expert Systems with Applications 41, no. 5 (2014): 2463-2471."
DOI for the article
Management and Information Systems