Image Annotation by Learning Rules from Regions Patterns

The modeling of multimedia and especially the semantic gap between the visual features and semantic concepts become an important domain due to the quantity of visual digital content, which speedily grows. In this paper, the analysis and semantic annotation of images are studied. The development of methods for colour image annotations based on learning represents the main contribution of the paper. The developed algorithms generate semantic pattern rules that identify high-level image concepts. A semantic pattern rule is a combination of images´ region patterns that identifies semantic concepts. Our methods are not limited to any specific domain and they can be applied in any field.