||Single channel object representations are often insufficient in modern image analysis applications;
novel methods which can handle heterogeneous information-rich multi-channel data are needed.
Such representations can incorporate information about a variety of different object properties.
They may result from fusion of information from different imaging modalities, but also by considering several layers of computed local image features.
In our recent paper we suggest to model each property by a fuzzy set over the image domain, i.e., we create a vector valued fuzzy set to store the multi-channel representation. We introduce the concept of a set distance measure applicable to such representations, extending some of our previously proposed set distances.
During the talk, I will present the idea and some promising results for template matching and object classification using the proposed distance measure.