Extended Statistical Landscape Features for Dynamic Texture Recognition

This paper proposes a new method for describing Dynamic Texture (DT). DT is an extension of still texture to temporal domain, which contains motion features and appearance features. An Extended Statistical Landscape Features (ESLF) method is proposed for DT description and recognition by characterizing the motion and appearance features. The proposed ESLF uses the ESLF histogram as the identifier of DT, which is concatenated by the local motion pattern (LMP) histogram derived from motion features and the SLF histogram from appearance features. Experimental results based on the DynTex database show that the proposed ESLF achieves a higher recognition performance than LBP-TOP.