| Summary: | 博士 === 國立中央大學 === 資訊工程學系 === 104 === Changing dynamic texture appearances can create new looks in both motion and color appearance of videos. For instance, waterfall texture in a video scene can be appeared as a fire texture or vice versa. This dissertation proposes a novel method for dynamic texture transformation between two dynamic textures by strategic motion coherence analysis.
Dynamic textures with sophisticated shape and motion appearance are difficult to represent by physical models and are hard to predict, especially for transformation to new motion texture. This study proposes dynamic texture transformation algorithms for video sequences based on the difference of pixel intensity and motion coherence of patches. This research successfully applies the technology in many special effect videos, using an interactive tool developed. Our study addresses the issues of 3D patch creation, motion coherence analysis and patch matching for dynamic texture transformation. The main contribution includes two issues. First is to propose a new metric for evaluating motion coherence, with solid tests to justify the usefulness (close to human eye perception). The second is to propose algorithms for automatic dynamic texture transformation. An optimized algorithm only needs users to target the texture on the first frame, by using an optional threshold to determine the texture area. The rest process is complete automatically by our proposed algorithms. The experimental results show that the motion coherence index is effectively used to find the coherent motion region for patch matching and transformation.
Besides, the distinct of motion coherence for each dynamic texture is also observed and analyzed to discuss in this research. A contribution found from the distinct of motion coherence is possibly aimed to leverage the motion coherence for other system development. For instance, the digital security system of next generation by improving strong point and significantly complementing defects of the existing closed circuit television (CCTV) system, by integrating motion analysis module into the existing surveillance system.
|
|---|
