A Combination of Machine Learning and Image Processing Technologies for the Classification of Image Regions

If large amounts of images are available, as it is the case in image archives, image retrieval technologies are necessary to help the user to find needed information. In order to make such queries possible, images have to be enriched with content-based annotations. As manual annotation is very costly, system support is desired. This paper introduces an approach to learning image region classifiers from extracted color, texture, shape, and position features. In different experiments, three machine learning algorithms were applied for the classification of character shapes and regions in landscape images.