Automatic animation of high resolution images

In the report, we present a novel photo-animation method and an animation tool for transformation of a photo into an animated video clip. In opposite to the existing tools, our technology provides both automatic and intuitive interactive photo-animation of high-resolution images. Our method and technology comprise the following main components: 1. Vectorization of the image, i.e. its representation by the sets of primitive structures (geometric models) in different scales. 2. 3D-2D global scale kinematic models for various image objects, especially, for the patterns of human bodies. These models allow an adaptation to definite parameters of the patterns and provide a high visual quality of motions' reproduction, including relatively large 3D rotations. 3. Tools and methods for translation of true 3D motion (like various scenarios of human 3D motion appearing in the Carnegie Mellon University database) into 2.5D animation. 4. Tools and methods for real-time automatic fitting of the models to image objects. In the course of fitting, specific anatomic parameters of the pattern are detected and preserved. The suggested method starts with face detection and model fitting, and then extends up to a full body pose capturing and model fitting. Our approach strongly relies on information provided by a vectorized image data, like the geometry of the edges and ridges and their color profiles, and on the transformations of 3D and 2D models. 5. High level scenarios and their automatic adaptation to the actual position of the characters in the photo. The presented technology is illustrated by several examples. Links to more running examples and to our on-line Photoanimation tool are provided.