Search

E-government offers opportunities for improving the interactions between citizens, governmental and public institutions, private sector organizations, and public employees. Despite this, the take-u...

Most supervised classification algorithms produce a soft score (either a probability, a fuzzy degree, a possibility, a cost, etc.) assessing the strength of the association between items and classes. After that, each item is assigned to the class with the highest soft score. In this paper, we show that this last step can be improved through alternative procedures more sensible to the available soft information. To this aim, we propose a general fuzzy bipolar approach that enables learning how to take advantage of the soft information provided by many classification algorithms in order to enhance the generalization power and accuracy of the classifiers. To show the suitability of the proposed approach, we also present some computational experiences for binary classification problems, in which its application to some well-known classifiers as random forest, classification trees and neural networks produces a statistically significant improvement in the performance of the classifiers.

Traditionally, the last step of edge detection algorithms, which is called scaling-evaluation, produces the final output classifying each pixel as edge or nonedge. This last step is usually done based on local evaluation methods. The local evaluation makes this classification based on measures obtained for every pixel. By contrast, in this work, we propose a global evaluation approach based on the idea of edge list to produce a solution that suits more with the human perception. In particular, we propose a new evaluation method that can be combined with any classical edge detection algorithm in an easy way to produce a novel edge detection algorithm. The new global evaluation method is divided in four steps: in first place we build the edge lists, that we have called edge segments. In second place we extract the characteristics associated to each segment: length, intensity, location, and so on. In the third step we learn the characteristics that make a segment good enough to become an edge. At the fourth step, we apply the classification task. In this work we have built the ground truth of edge list necessary for the supervised classification. Finally, we test the effectiveness of this algorithm against other classical algorithms based on local evaluation approach.

Most edge detection algorithms deal only with grayscale images, and the way of adapting them to use with RGB images is an open problem. In this work, we explore different ways of aggregating the color information of a RGB image for edges extraction, and this is made by means of well-known edge detection algorithms. In this research, it is been used the set of images from Berkeley. In order to evaluate the algorithm’s performance, F measure is computed. The way that color information -the different channels- is aggregated is proved to be relevant for the edge detection task. Moreover, post-aggregation of channels performed significatively better than the classic approach (pre-aggregation of channels).

Edge detection problems try to identify those pixels that represent the boundaries of the objects in an image. The process for getting a solution is usually organized in several steps, producing at the end a set of pixels that could be edges (candidates to be edges). These pixels are then classified based on some local evaluation method, taking into account the measurements obtained in each pixel. In this paper, we propose a global evaluation method based on the idea of edge list to produce a solution. In particular, we propose an algorithm divided in four steps: in first place we build the edge list (that we have called segments); in second place we extract the characteristics associated to each segment (length, intensity, location,…); in the third step we learn which are the characteristics that make a segment good enough to be a boundary; finally, in the fourth place, we apply the classification task. In this work we have built the ground truth of edge list necessary for the supervised classification. Finally we test the effectiveness of this algorithm against other classical approaches.