Frequency-space Analysis of Cervical Images using Short Time Fourier Transform

V. Van Raad (Australia)


Frequencyspace analysis, cervical images, STFT,cervical cancer prevention, Gaussian mixture model,transformation zone, feature space


In colposcopy systems, used for cervical cancer prevention, visualisation and treatment, an important task is to evaluate the state of the cervix "in vivo". In the presented work we propose a frequency-space analysis of cervical images using a two dimensional Short Time Fourier Transform (STFT). We applied a Gaussian Model Mixture-based method to correct the cervical images for an artefact, known as Specular Reflection (SR). Further, we analysed the pre-processed grey scale images in three frequency bands within the frequency interval [0,π/2]and we outlined cervical features such as the transformation zone (TZ) and squamo-inter-epithelial lesions (SIL). These are presented as tri-colour (pseudo) feature images, eventually used for diagnostic aid in colposcopy. We establish important frequency estimates for modelling and segmentation.

Important Links:

Go Back