Abstract: Chromosome classification is very significant in the domain of medical image processing, and feature extraction is the key within this process. In this paper, the topographic analysis approach to extract chromosome features in gray-level image, which is operated by scale-space filter, is described. Bands on chromosome are described as a series of peaks and saddles, and symmetric axes as ridges. Differential geometry is used for the exact definition of ridge, as well as peak and saddle. The results of feature extraction in terms of those definitions are also presented, and they seem more precise for the classification task.