The main aim of the project here presented is the design of a new method for repetitive (structural) texture recognition and modelization. This method must be able to solve a series of situations such as overlapping of a texture on an image of variable background, the existence of nestings (hierarchies) in the textural elements, frequency and phase variation in the repetition, shape, size and contrast variations of the texture elements, etc. Most of the proposed models assume that these situations do not arise; however, they are frequent in real scene images. Our approach consists of constructing a model from the transformations that a texture suffers when examined at different scale and detail levels. In order to achieve this, we begin by studying the properties and results of different "multi-scale analysis" techniques: space-scale methods, multi-resolution pyramides, decomposition by
|Effective start/end date||15/06/92 → 15/06/95|
Explore the research topics touched on by this project. These labels are generated based on the underlying awards/grants. Together they form a unique fingerprint.