Facial Feature Extraction Technology And System For Cartoon Animation

2D facial cartoon animation is the most important part of the 2D cartoon animation, but also is the difficult part. Viewpoint and stroke information is the important information of facial cartoon Hand-drawing,which is determined before animatior finishes it. But in process of computer-aided animation generation, computer does not know these information, it requires a combination of prior knowledge to "inference" viewpoint and stroke information of hand-drawing.In this paper, we will focus on the recovery of hand-drawing information,viewpoint and stroke. At the same time,we will finish a project of automatic recognition of viewpoint and stroke information, and user can also correct recognition errors of viewpoint and stroke information.To achieve these goals, we need to complete the following tasks:1) Generating samples of machine learning automatically, we introduce the three-dimensional approximated model to generate samples.2)Viewpoint recognition. Using support vector machine (SVM), the samples are classified by the classification model to identify the viewpoint of hand-drawing.3) Hand-drawing strokes recognition. We also use the SVM classifier to identify stroke information.4) Automatic feature selection. reduce the dimension of feature vector, remove redundant features to achieve the goal of higher precision and efficiency of machine learning.