Computer Assisted Hand-drawn Animation Of Faces And Cloth

Hand-drawn animation is one of the most infectious form of traditional two-dimensional animation. With the development of computer technology, the production of hand-drawn animation is no longer confined to the digital process of traditional animation, a growing number of studies trying to achieve the automatic generation of hand-drawn animation, to reduce labor and speed up the production process. In recent years, computer-assisted hand-drawn animation becomes a research focus in the fields of computer graphics, computer animation, computer vision and other related areas. As the modeling, rendering and motion of hand-drawn characters are tightly coupled in 2D frame sequences of which the essential part is composed of key frames, computer assisted traditional 2D hand-drawn animation is still facing many difficulties compared with three-dimensional animation.To bridge the gap, in the paper, we study the technique of computer assisted hand-drawn animation, focusing on the animation of hand-drawn cartoon faces and hand-drawn cloth, which is the most expressive and time-consuming. Specifically, the main research topics of this thesis are as follows:1. We expanded existing performance-driven animation techniques to generate multi-view hand-drawn facial animation. A 3D approximate facial model automatically built from front and side view master frames of character drawings is introduced to enable the animated cartoon faces to be viewed from angles different from that in the input video, and the lit sphere, capturing the lighting in the painting artwork of faces, is utilized to color the cartoon faces in terms of the 3D approximate facial model, reinforcing the hand-drawn appearance of the resulting facial animation.2. Research on the generation of exaggerated facial animation based on Learning from examples. The expressive mappings are built by artificial neural network (ANN) trained from the examples of the real face in the video and the cartoon facial drawings in the facial expression graph for a specific character. The learned mapping model makes the resultant facial animation to properly get the desired expressiveness, instead of a mere reproduction of the facial actions in the input video sequence. The system "infers" the facial movements of a cartoon character by an expressive mapping model trained from a set of examples of key expressions performed in the video and the pre-defined representative expressions constructed in a facial expression graph for a specific character, makes the resultant cartoon faces go beyond the real life rather than just mimic it, which can visually preserve the artistic style of animation and better convey the personality of a cartoon character.3. We studied the generation of hand-drawn cloth animation with 3D information, to keep the proportion of cloth in computer assisted hand-drawn animation, and preserve the movement of cloth in real life. According to 3D approximate cloth model and hand-drawn style, multi-view hand-drawn cloth animation of any pose can be obtained by motion data.3D geometric information, realistic physics simulation, and 2D hand-drawn styles, are combined together with no extra modeling work, to automatically generate authentic hand-drawn cloth animation.4. Proposed a novel way to establish the dynamic model of cloth movement from 2D cloth drawings. The dynamic model of cloth movement reflects the hand-drawn style and physical movements of cloth completely. It matches the input key-frame 2D drawings, and reflects the animator's understanding of the cloth movement. GPU parallel computing is employed to speed up the time-consuming physical simulation of cloth motion for the calculation of kinetic parameters. Under the guidance of the dynamic model of cloth movements, we can produce hand-drawn cloth animation sequences which convey the physical principles of cloth motion while well-preserving hand-drawn styles.5. Proposed a novel way to generate hand-drawn clothing animation by 2D drawings and the physical laws of cloth movements. It firstly recovers the kinetic parameters of cloth motion inside the 2D drawings, and then by the principles of traditional key-frame animation, the in-between frames are automatically created by physical simulation based on interpolating the kinetic parameters from key frames. Compared with the traditional interpolation method, our method generates the hand-drawn animation of physical movements of cloth. 6. Learn from 3D animation method, we proposed a novel way to generate hand-drawn clothing animation by a single reference hand-drawn frame, and extrapolating the kinetic parameters recovered from reference frame according to the specified motion capture data. Compared with non-photorealistic rendering methods based 3D model, our method based on the deformation of hand drawings, which retaining the exaggeration and artistry of 2D animation.