Automatic classification of skin lesions using geometrical measurements of adaptive neighborhoods and local binary patterns

V. Gonzalez-Castro, J. Debayle*, Y. Wazaefi, M. Rahim, C. Gaudy, J. J. Grob, B. Fertil

*Corresponding author for this work

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

This paper introduces a method for characterizing and classifying skin lesions in dermoscopic color images with the goal of detecting which ones are melanoma (cancerous lesions). The images are described by means of the Local Binary Patterns (LBPs) computed on geometrical feature maps of each color component of the image. These maps are extracted from geometrical measurements of the General Adaptive Neighborhoods (GAN) of the pixels. The GAN of a pixel is a region surrounding it and fitting its local image spatial structure. The performance of the proposed texture descriptor has been evaluated by means of an Artificial Neural Network, and it has been compared with the classical LBPs. Experimental results using ROC curves show that the GAN-based method outperforms the classical one and the dermatologists' predictions.

Original languageEnglish
Title of host publicationProceedings - International Conference on Image Processing, ICIP
PublisherInstitute of Electrical and Electronics Engineers (IEEE)
Pages1722-1726
Number of pages5
Volume2015-December
ISBN (Print)9781479983391
DOIs
Publication statusPublished - 9 Dec 2015
EventIEEE International Conference on Image Processing, ICIP 2015 - Quebec City, Canada
Duration: 27 Sep 201530 Sep 2015

Conference

ConferenceIEEE International Conference on Image Processing, ICIP 2015
Country/TerritoryCanada
CityQuebec City
Period27/09/1530/09/15

Keywords

  • General adaptive neighborhoods
  • Local binary patterns
  • Melanoma
  • Texture description

Fingerprint

Dive into the research topics of 'Automatic classification of skin lesions using geometrical measurements of adaptive neighborhoods and local binary patterns'. Together they form a unique fingerprint.

Cite this