R. Dobrescu, F. Talos, M. Dobrescu, and D. Andone (Romania
fractal dimension, textural features, image retrieval, similarity, histogram method
The objective of the present work is to offer an alternate solution for the evaluation of the textural features that characterize different types of material. In this aim, an algorithm that compute the fractal dimension of a wall painting bitmap image was used to determine information on the boundary between different zones of the images with same textural aspect. A classification system prototype was built for evaluation purposes, taken in contest other algorithms that use Haralick's textural features. The use of Principal Component Analysis to reduce the redundancy in the Haralick's textural features allowed an expressive improvement in the classification performance. Then, the possibility to use textural features in image retrieval from a database and a procedure to take decisions on the image relevance are presented.
Important Links:
Go Back
