A Method for Assessment of Segmentation Success Considering Uncertainty in the Edge Positions
- PDF / 3,364,055 Bytes
- 12 Pages / 600.03 x 792 pts Page_size
- 15 Downloads / 148 Views
A Method for Assessment of Segmentation Success Considering Uncertainty in the Edge Positions ´ Usamentiaga, Daniel F. Garc´ıa, Carlos Lopez, ´ ´ Ruben and Diego Gonzalez Department of Computer Science, University of Oviedo, Campus de Viesques, 33204 Gij´on, Asturias, Spain Received 27 February 2005; Revised 6 June 2005; Accepted 27 June 2005 A method for segmentation assessment is proposed. The technique is based on a comparison of the segmentation produced by an algorithm with an ideal segmentation. The procedure to obtain the ideal segmentation is described in detail. Uncertainty regarding the edge positions is accounted for in the discrepancy calculation of each edge using fuzzy reasoning. The uncertainty measurement consists of a generalization, using fuzzy membership functions, of the similarity metrics used by well-known assessment methods. Several alternatives for the fuzzy membership functions, based on statistical properties of the possible positions of each edge, are defined. The proposed uncertainty measurement can be easily applied to other well-known methods. Finally, the segmentation assessment method is used to determine the best segmentation algorithm for thermographic images, and also to tune the optimum parameters of each algorithm. Copyright © 2006 Rub´en Usamentiaga et al. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.
1.
INTRODUCTION
Image segmentation is one of the most important components in an image analysis system. The objective of segmentation is to divide the image into meaningful regions. After the segmentation, features of each region are identified to be used for further analysis. Since the analysis of the image is based on the identified features, and the features are calculated from the segmented regions, the accuracy of the segmentation is crucial to the performance of the image analysis system. Over the last few decades, many segmentation algorithms have been proposed [1]. However, the evaluation of the performance of these algorithms is usually poor, consisting of the presentation of a few segmented images. In order to evaluate segmentation algorithms, several evaluation methods which can be used to determine the effectiveness of an algorithm, and which also allow the comparison of several algorithms, have been proposed. In this work a new segmentation assessment method which does not have the well-known problems of other methods, and which also takes the uncertainty into account, is proposed. This method will be used to decide which algorithm is the best for the segmentation of thermographic images and also to find the optimum parameters of each algorithm.
2.
PREVIOUS EVALUATION METHODS
Zhang [2] proposes a classification of existing assessment methods as “analytical,” “empirical goodness,” and “empirical discrepancy.” Other authors, such as Yang et al.[3], use a different classification: “supervised” and “uns
Data Loading...
