Luca Costantini

Learn More
Action recognition in videos is a relevant and challenging task of automatic semantic video analysis. Most successful approaches exploit local space-time descriptors. These descriptors are usually carefully engineered in order to obtain feature invariance to photometric and geometric variations. The main drawback of space-time descriptors is high(More)
In this work the authors propose a classification method based on Support Vector Machine (SVM) and key frames features extraction to classify historical sport video contents. In the context of the Italian Project, IRMA (Information Retrieval in Multimedia Archives), with the goal to recover and preserve historical videos of proven cultural interest, a data(More)
In this work a novel technique for representing the edges of an image is presented and the impact of this on image clustering is investigated. The characterization is performed in two steps: the “most important” edges are first selected by using both the Laplace operator and the Laguerre Gauss functions, and then the phase distribution of each(More)
A novel technique via Quadtree Gauss-Laguerre Transform (LGT) for a complete localization of a complex image in an multimedia database is presented. This technique is based on an iterative Maximum Likelihood procedure that allows to compare a region of interest (ROI) of an image with a content of a database, independently from location, rotation and scale.(More)
We estimate sentiment categories proportions for retrieval within large retrieval sets. In general, estimates are produced by counting the classification outcomes and then by adjusting such category sizes taking into account misclassification error matrix. However, both the accuracy of the classifier and the precision of the retrieval produce a large number(More)