Synthesizing And Editing Flowing Water And Fire Using Video

The natural landscape simulation is an important research area of computer graphics. Existing simulation techniques mainly are fractal geometry technology and particle system. Based on video texture synthesis technique, we propose a new approach for the synthesis of flowing water and fire video.Video texture is a new median introduced in recent years, which is something inbetween the photo and video that captures dynamics of the video and looks realistic. To synthesize a video with video texture the temporal continuity must be maintained. We can synthesize a continuous stream of images by repeating loops detected in the original video.We investigate the methods of synthesizing and the flowing water and fire. In the case of synthesizing and editing flowing water, we first run the video stabilization and loop detection, we then extend the video textures using a Markov Random Field to fit the required length of the flowing water in the scene, we can further edit the video textures such as changing the stone sizes or even removing stones from the video textures. The flow continuity is preserved well in the resultant water animations.In the process of synthesizing fire video, we adopt different image segmentation methodologies for different fire images, and then edit the fire textures to fit scenes. Finally we build simple models to simulate the wind effect and fire spreading process, as demonstrated by examples in the thesis.