In the field of texture analysis we investigate the classification and segmentation of textures. Textures are characterised by a random visual structure that does not show a particular object. The research focus is on feature-based description of textures with a particular emphasis on features that are related to the processing steps in the human visual cortex.

Fault Detection

In the field of quality control faults on textured surfaces need to be detected and evaluated. Quite often the detection has to reproduced human cognitive abilities to extract only those areas that are considered disturbing by a human. The research approach is on the one hand a classical feature/classifier approach and on the other hand a more flexible approach that tries to automatically construct whole segmentation algorithms (rather than just parameter tuning).

Judging Aesthetic Properties

The design of a surface quite often includes visual textures. The design process is often only based on the subjective assessment of the designer, probably accompanied by panel test. Our research focuses on computational models that allow an accurate estimation of the aesthetic properties of visual textures. Based on such predictive models we develop texture synthesis methods are used to create texture for a given set of aesthetic properties.