Component trees in Milena

From LRDE

Abstract

Pattern recognition and object detection are very important stakes in image processing. Many solutions have been provided; nevertheless the one based on component trees seems to be the most promising. A component tree allows to work on connected components at different gray-levels in the image. Thanks to this tree, we can use attributes as criteria to identify components, such a component being an object in the image. During this seminar, we present a component tree implementation, how to deal with attributes, different methods dedicated to component identification and a general processing chain that leads to object recognition.