Seminar Series

Multifractal Techniques for Image Analysis and Classification with Applications to Emphysema Databases


Musibau Ibrahim


The University of Canterbury

Time & Place

Mon, 26 Dec 2016 09:00:00 NZDT in Erskine 315.


Statistical measures of self-similarity that represent the fractal nature of intensity distributions have been widely used in the bio-medical image processing. This seminar presents novel multi-fractal based approaches for the classification of Emphysema patterns in HRCT images. The talk begins with an overview of important computational aspects related to the estimation of fractal dimensions and the Holder exponent (α-value) that represents the variation of intensities in local pixel neighbourhoods. The fractal dimension corresponding to each α-value gives a multi-fractal spectrum from which feature descriptors could be constructed for classification. These feature vectors could be effectively combined with other attributes of global intensity distributions such as alpha-histograms and local binary patterns to design classifiers for the Emphysema dataset. The seminar also presents feature selection techniques that have shown to increase the discriminating capability of the feature vectors and improve classification accuracy in our experiments.


Ibrahim M. Adekunle received his B.Sc. degree in Computer Science from University of Ilorin in Nigeria, and M.Sc. in Computer and Network Engineering from Sheffield Hallam University in the United Kingdom. Recently, he has been pursuing a PhD degree in the Department of Computer Science and Software Engineering at the University of Canterbury. His research interests are in the field of biomedical image analysis and classification.