Unsupervised Object Matching and Categorization via Agglomerative Correspondence Clustering

This paper presents computationally efficient object detection, matching and categorization viaAgglomerative Correspondence Clustering (ACC). We implement ACC for feature correspondence andobject-based image matching exploiting both photometric similarity and geometric consistency from localinvariant features. Object- based image matching is formulated here as an unsupervised multi-classclustering problem on a set of candidate feature matches linking maximally stable external regions featuresand scale invariant features in the framework of hierarchical agglomerative clustering. The algorithmiscapable to handle significant amount of outliers and deformations such as scaling and rotation as wellasmultiple clusters, thus powering simultaneous feature matching and clustering from real-world image pairswith significant clutter and multiple objects. Theexperimental assessment on feature correspondence,object recognition, and object- based image matching demonstrates that, this method is capable ofrigorously handling scaling, rotation, and deformation and can be applied to a wide range of imagematching and object recognition and categorizationrelated real-world problems.