Image Based Synthesis Of Fire

Computer simulation of fire is an important research area. Existing fire simulation techniques can be classified into two categories: direct numerical simulation and visual modeling. Base on texture synthesis technique developed in recent years, we propose a new approach for the synthesis of fire.Computer simulation of fire needs to solve two problems: determination of fire contour and fire texture inside the contour. In this thesis we deal with fire contour by use of noise function or abstracting the contour from real fire series.To synthesize fire we need to following problems: 1. maintaining the movement continuity of varying fire texture in the synthesized fire series; 2. natural transition between textures from the central part to the boundary part of the fire; 3. processing of overlap parts in the process of fire spreading.We divide the fire sample image into several levels vertically and the synthesis process is carried out by searching candidates on the corresponding levels. Textures on the central and boundary parts are synthesized separately, and natural transition between them is achieved by use of Markov probability interpolation. The texture on the overlap area between neighboring fire bodies is dealt with by use of linear interpolation of textures between crossing areas of the two fire bodies. Finally we show how to simulate large fire body in 3D space and discussed future research work based on current approach.