| Summary: | 碩士 === 國立成功大學 === 機械工程學系碩博士班 === 95 === Human head reconstruction becomes an important research topic while the computer graphic technologies developing in past few decades. Because of large amount of the face features are complex, and the needs of the real-time animation of the facial expressions, it is necessary to elaborate the head model. In the past, researchers often selected the features by hands. It is a subjective method. This research uses an objective and automatic method to locate the features on the head. The reconstruction of head model is according to the feature points and lines, and provides different levels of details to fit different requirements. All of these levels of meshes will not lose the features.
This article improves the method described in “Feature-based Digital Head Reconstruction.” Systematically and objectively extract features automatically according to the MPEG-4 definition. This research also introduces a method that can rectify the tilt head to enhance the recognition, and a method that can replace the poorly sampled ear data from the body scanner by a better one from the CT image. The extendable feature-based head model can be easily changed the density of the meshes according to the requirement. It is much better suitable for the applications of data transmission across the internet and computer graphics animation.
|
|---|
