Affiliation:
1. University of Canterbury, New Zealand
Abstract
Tissue image classification is a challenging problem due to the fact that the images contain highly irregular shapes in complex spatial arrangement. The multi-fractal formalism has been found useful in characterizing the intensity distribution present in such images, as it can effectively resolve local densities and also represent various structures present in the image. This paper presents a detailed study of feature vectors derived from the distribution of Holder exponents and the geometrical characteristics of the multi-fractal spectra that can be used in applications requiring image classification and retrieval. The paper also gives the results of experimental analysis performed using a tissue image database and demonstrates the effectiveness of the proposed multi-fractal-based descriptors in tissue image classification and retrieval. Implementation aspects that need to be considered for improving classification accuracy and the feature representation capability of the proposed descriptors are also outlined.
Reference32 articles.
1. Aksoy, S., Marchisio, G., Tusk, C., & Koperski, K. (2002). Interactive classification and content-based retrieval of Tissue Images. In Proceedings of the SPIE Annual Meeting, Applications of Digital Image Processing Session (Vol. 4790, pp. 71-81).
2. Random Self-Similar. Multi-fractals;M.Arbeiter;Mathematische Nachrichten,1996
3. Geometry of multifractal systems
4. Direct determination of the f(α) singularity spectrum
5. Esgiar, A. N., & Chakravorty, P. K. (2007). Fractal based classification of colon cancer tissue images. In Proceedings of the International Symposium on Signal Processing and its Applications (pp. 1-4).