Vitiligo is a pathology that causes the appearance of macules achromic (white spots) in the skin. Besides, generates a negative emotional burden in the people that have it, what make necessary to develop suitable methods to identify and treat it properly. In this paper we propose a novel system formed by two stages: The Front End where the principal characteristics of the image are extracted using the Mel Frequency Cepstral Coefficients (MFCC) and i-Vectors (techniques widely used in speech processing) and the Back End, where these characteristics are received and through a classifier is define whether and image contains or not vitiligo. Artificial Neural Networks and Support Vector Machines were selected as classifiers. Results shows that both MFCC and i-Vectors could be used in the field of image processing. Although, the i-Vectors allows us to decrease more the dimensionality of a feature vector and without losing the characteristics of the high dimensionality, this was reflected in their performance with an accuracy of 95.28% to recognize correctly images.
CITATION STYLE
Salamea, C., & Chica, J. F. (2020). Vitiligo Detection Using Cepstral Coefficients. In Advances in Intelligent Systems and Computing (Vol. 1066, pp. 389–398). Springer Science and Business Media Deutschland GmbH. https://doi.org/10.1007/978-3-030-32022-5_36
Mendeley helps you to discover research relevant for your work.