Special issue on dynamic textures in video

Abstract

Two-dimensional textures in images have been extensively studied in the past. On the other hand, there is comparatively limited research on three-dimensional dynamic textures that exhibit certain time-varying properties in video. In many scenes there are regions having significant structural similarity and exhibit high temporal correlations between image frames forming the video [1]. A tree swaying in the wind or a wave lapping on a beach is not just a collection of randomly shuffled appearances, but a physical system that has characteristic responses associated with its dynamics. Examples of such dynamic phenomena include flames, smoke, sea, waves, clouds, fog, crowds in public places and sports events, some human movements, and even shadows [1–9]. It is known that dynamic textures, especially for outdoor scenes, cause major problems in motion detection and analysis tasks. Besides, they drastically decrease the coding efficiency of video encoders although they do not contain any useful and discriminative information. They complicate motion-based object recognition methods. By segmenting and excluding dynamic textures, the robustness of the moving object detection and action identification can be improved. Other practical applications include detection of certain types of dynamic textures such as fire and smoke, realistic rendering and compact visualization of dynamic textures, and efficient retrieval of video in multimedia databases. The