Rotationally invariant texture classification

Texture based features used for content based retrieval of images and videos should ideally be invariant to simple transforms such as rotation. This paper introduces the recently developed dual tree complex wavelet transform (DT-CWT) as a tool to extract rotationally invariant texture based features. When applied in two dimensions the DT-CWT produces shift invariant and orientated subbands at each decomposition scale. Rotationally invariant features can be extracted from the energies of these subbands whilst benefiting from the computational efficiency of the decomposition and the ability to choose the transform filters