Contextual object categorisation with energy‐based model

In image mining, the object categorisation is an important issue. In the previous researches for an object categorisation, the contextual information, especially the spatial relations between objects, has not been made full use. In addition, the object categorisation methods, which generally use the probabilistic graphical models to implement the incorporation of contextual information with appearance of objects, are almost inevitable to evaluate the intractable partition function for normalisation. In order to utilise fully the spatial relational information the authors introduced fully connected fuzzy spatial relations including general spatial relations between object regions. Moreover, then the spatial relations were considered as well as co‐occurrence and appearance of objects by using energy‐based model that is based on the region–object association potential and the configuration potential of objects. Minimising the energy function of whole image arrangement, they obtained the optimal label set about the image regions and addressed the evaluation of intractable partition function in conditional random fields.