The design and implementation of realistic fish animation play system using video sequence

In this paper, a novel system for rendering and animating fish is proposed. We captured a short video sequence of fish, and analyzed the similarities between these fish in video frames. In order to control the motions of the fish interactively as in computer games, we attempted to learn the motion properties of the fish in this video sequence. Some of the motion parameters of the fish were extracted from the video frames and then trained using a neural network. This trained information provided motion control for generating fish animation. In the proposed system, we also generate arbitrary video loop lengths for rendering and animating fish. It is not necessary that the video samples in these video loops be played in sequence. Furthermore, we can adapt both the video loop and neural network techniques in order to control the fish motions interactively and to maintain visual smoothness in fish animation.