Kronecker decomposition for image classification

Abstract

We propose an image decomposition technique that captures the structure of a scene. An image is decomposed into a matrix that represents the adjacency between the elements of the image and their distance. Images decomposed this way are then classified using a maximum margin regression (MMR) approach where the normal vector of the separating hyperplane maps the input feature vectors into the outputs vectors. Multiclass and multilabel classification are native to MMR, unlike other more classical maximum margin approaches, like SVM. We have tested our approach with the ImageCLEF 2015 multi-label classification task, Pascal VOC and Flickr dataset.