Normal versus complete flow in dynamic texture recognition: a comparative study

We address the problem of dynamic texture (DT) classification using different techniques based on optic flow. The optic flow based approaches dominate among the currently available DT classification methods [4]. The optic flow features used by these methods often describe the local image distortion in terms of such quantities as curl or divergence. Both normal and complete flow have been considered, with the normal flow being used much more frequently. However, the precise meaning and the applicability of the normal and the complete flow features have never been analysed properly. In this paper, we provide a principled analysis of local image distortions and present the results of a DT classification study that compares the performances of the two types of flow with different features. The effect of the flow confidence measure introduced in [8] is also discussed.