Recent advances in image analysis and pattern recognition have paved the way to several developments in plant science. In the present work, we report the comparative study, by using the aforementioned approaches, of vascular bundles of Dracaena marginata. More specifically, we used 33 measurements related to shape, density and regularity of imaged cross-sections of the stem. By using individual, pairwise and PCA projections of the adopted measurements, we were able to find the combinations of measurements leading to the best separation between the considered tissues. In particular, the best separation was obtained for entropy taken at a particular spatial scale combined with the equivalent diameter. The reported developments open several perspectives for applications in content-based retrieval, diagnosis, and species identification.