Point Pattern Matching with Locality Preserving Descriptors

This paper proposes a novel feature descriptor - locality preserving descriptor to the problem of point pattern matching. The idea behind a locality preserving is to map points that are nearby in the data space into points that are nearby in the feature space. The feature descriptor optimally preserves the neighborhood structure of the data set, and is invariant to translation, scale, and rotation. We use of the locality preserving descriptor and combine the continuity constraint for point pattern matching.