Classification of color colposcopic images by neural networks

This article presents colposcopic image classification based on color parameters used in a comparison study of different artificial neural networks and the k-nearest neighbors reference method. In this study, significant image data bases are used (617 images) from which two sets of parameters is extracted to characterize the attribute of color. More precisely, we select a set of color componants from color spaces based on data analysis and inner characteristics of different colposcopic image examinations. Experimental results show the feasibility of this study and the efficiency of the two sets of parameters since 90.9% of pink/red image set and 78% of brown/yellow image set have been correctly classified.