Video texture synthesis : A survey

A video texture is nothing but a continuous, infinitely varying stream of video images. It is a sequence of images with visual pattern repetition in spatial and time domain, like river flow, smoke, fire, clouds, water fall, and windy grass field etc. The simplest way of producing a video texture is to synthesize it from a finite set of images by randomly rearranging (and possibly blending) original frames from source video which plays a vital role in computer vision and computer graphics. The algorithms for synthesizing video textures that have proposed over the past one decade have been surveyed here. A variety of algorithms associated with synthesized video textures are analyzed. There are two incompatible goals for video texture synthesis that need to be accomplished: (a) the synthesized and the input texture have visual similarity and (b) temporal coherence and motion consistency or evolution of the video texture. The video texture generated must be in desired flow with respect to time.