Please use this identifier to cite or link to this item:
http://acervodigital.unesp.br/handle/11449/127107
- Title:
- Fractal descriptors for discrimination of microscopy images of plant leaves
- Universidade Estadual Paulista (UNESP)
- Universidade de São Paulo (USP)
- Universidad Nacional del Litoral
- 1742-6596
- This study proposes the application of fractal descriptors method to the discrimination of microscopy images of plant leaves. Fractal descriptors have demonstrated to be a powerful discriminative method in image analysis, mainly for the discrimination of natural objects. In fact, these descriptors express the spatial arrangement of pixels inside the texture under different scales and such arrangements are directly related to physical properties inherent to the material depicted in the image. Here, we employ the Bouligand-Minkowski descriptors. These are obtained by the dilation of a surface mapping the gray-level texture. The classification of the microscopy images is performed by the well-known Support Vector Machine (SVM) method and we compare the success rate with other literature texture analysis methods. The proposed method achieved a correctness rate of 89%, while the second best solution, the Co-occurrence descriptors, yielded only 78%. This clear advantage of fractal descriptors demonstrates the potential of such approach in the analysis of the plant microscopy images.
- 2014
- Journal of Physics. Conference Series, v. 490, n. 1, p. 12085, 2014.
- 12085
- http://iopscience.iop.org/1742-6596/490/1/012085/
- Acesso aberto
- outro
- http://repositorio.unesp.br/handle/11449/127107
There are no files associated with this item.
Items in DSpace are protected by copyright, with all rights reserved, unless otherwise indicated.