Dynamic texture recognition using normal flow and texture regularity.pdfVIP

223 Dynamic Texture Recognition Using Normal Flow and Texture Regularity Renaud P´eteri1 and Dmitry Chetverikov2 1 Centre for Mathematics and Computer Science (CWI), Kruislaan 413, 1098SJ Amsterdam, The Netherlands Renaud.PETERI@ 2 MTA SZTAKI - Hungarian Academy of Sciences 1111 Budapest, Kende u. 13-17, Hungary Abstract. The processing, description and recognition of dynamic (time-varying) textures are new exciting areas of texture analysis. Many real-world textures are dynamic textures whose retrieval from a video database should be based on both dynamic and static features. In this article, a method for extracting features revealing fundamental proper- ties of dynamic textures is presented. These features are based on the normal flow and on the texture regularity though the sequence. Their discriminative ability is then successfully demonstrated on a full classi- fication process. 1 Introduction 1.1 Context The amount of available digital images and videos for professional or private purposes is quickly growing. Extracting useful information from these data is a highly challenging problem and requires the design of efficient content-based retrieval algorithms. The current MPEG-7 standardization (also known as the “Multimedia Content Description Interface”) aims at providing a set of content descriptors of multimedia data such as videos. Among them, texture [12] and motion [4] were identified as key features for video interpretation. Combining texture and motion leads to a certain type of mo