Unsupervised Segmentation of Medical Images using DCT Coefficients

Singh, P.K.

    Image segmentation is a prerequisite process for image content understanding and visual object recognition in medical images for the development of a computer aided diagnosis(CAD) system. An unsupervised segmentation method is proposed which uses discrete cosine transform(DCT) coefficients for extraction of feature vectors and the Fisher Discriminant K-means (FDK) technique for clustering image pixels. In this study, the parenchymal region in HRCT lung images is separated first and then feature vectors using the deviation in local variance in DCT coefficients are determined for each pixels of parenchyma regions. The extracted feature vectors are used for selection of the best feature sets by reducing the dimensionality of the feature vector. The reduced feature vector is used for unsupervised classification using the K-means clustering algorithm which is guided by Fisher linear discriminant parameters for determining number of distinguishable regions in the image.
Cite as: Singh, P.K. (2004). Unsupervised Segmentation of Medical Images using DCT Coefficients. In Proc. 2003 Pan-Sydney Area Workshop on Visual Information Processing (VIP2003), Sydney, Australia. CRPIT, 36. Piccardi, M., Hintz, T., He, S., Huang, M. L. and Feng, D. D., Eds. ACS. 75-81.
pdf (from crpit.com) pdf (local if available) BibTeX EndNote GS