Abstract

ROA MARTINEZ, Sandra M.  and  MERA GAONA, Maritza F.. CONTENT-BASED RETRIEVAL OF CERVICAL MICROSCOPIC IMAGES INFECTED WITH HUMAN PAPILLOMAVIRUS USING TEXTURAL FEATURES. Rev. ing. biomed. [online]. 2013, vol.7, n.14, pp.69-80. ISSN 1909-9762.

The purpose of this work is to report the characterization process implemented to obtain visual descriptors representing visual content of digital images of cervical biopsies infected with Human Papilloma Virus (HPV). Positive biopsies with infected tissues present lesions known as Condyloma Plano Viral. A database of images of cervical biopsies was constructed in addition to the implementation of techniques that enhance the texture information and describe the content of images. This work proposed a set of features to describe the content of images from custom modifications of Discrete Wavelet Transform and the calculation of the Co-occurrence Matrix. This proposed feature set provided an average recovery rate of 80% in microscopic images of the cervix infected with HPV, from which CBIR systems have not been developed. Finally, this work determines the average recovery rate from the use of similarity metrics based on the standard LP.

Keywords : Feature extraction; Content Based Image Retrieval (CBIR); Tamura textures; Human Papilloma Virus (HPV).

        · abstract in Spanish | Portuguese     · text in Spanish     · Spanish ( pdf )