A gravitational model for grayscale texture classification applied to the pap-smear database

Abstract

This paper presents the application of a novel and very discriminative texture analysis method based on a gravitational model to a relevant medical problem, which is to classify pap-smear cell images. For this purpose, the complexity descriptors Bouligand-Minkowski fractal dimension and lacunarity were employed to extract signatures from the gravitational collapsing process. The obtained result was compared to other texture analysis methods. Additionally, AUC measure performance was computed and compared to several LBP based descriptors presented in two recent papers. The performed comparisons demonstrate that texture analysis based on gravitational model is suitable for discriminating pap-smear images.