10.1997 - 09.1999 ,    Код проекта: INTAS 96-785  Описание The necessity of multi-scale methods for image analysis has been recognised since the emergence of the field of image processing. Processing images at different scales has many benefits as it gives additional information about the essence of the percept which we want to understand. Today, there exists various alternative approaches: linear and nonlinear image scale-spaces, morphological imaging techniques, quadtrees, pyramids, wavelets, fractal-based techniques, etc. The goal of the current project was to contribute to this important research area, and more ambitiously, to explore their interrelationships. Furthermore, we wanted to apply such methods to various concrete problems, in particular the analysis of medical images (texture analysis, segmentation). An important application area where multi-scale methods can play an important role is Content-Based Image Retrieval (CBIR). Given the exponential growth of image and multi-media material in large databases and the Internet, there is increasing interest in search-engines that are able to retrieve images on the basis of image-oriented queries and visual clues. (Queries like: "Find images of deserted tropical beach'' or "Find images that look like this one''). As compared to more traditionally oriented databases (where one looks for an exact match with the query-item), image databases are searched for objects similar to the query-item. Research has shown that more advanced techniques for automatic image analysis and more sophisticated ways of shape recognition will be necessary to perform such tasks.  Ключевые слова segmentation, structure, texture, image processing, content-based image retrieval