Research On Face Cartoonlization Method For Video

Animated cartoon industry has experienced explosive growth in recent years. Cartoon characters are creatively applied to the video calls, role play, and so on, and it is inseparable from the computer aided animation synthesis technology, which plays an important role to reduce the workload and improve the visual effect. As to video face cartoonlization, there are two aspects have to be considered. Firstly, the processing rate must be fast enough to achieve the goal of real time. Secondly, the change of facial expression is subtle but important. So the video face cartoonlization has always been the emphasis and difficulty in computer aided animation synthesis technology.The existing methods of video face cartoonlization need complex pre-treatment or post a lot of work, and it’s difficult to generate personalized cartoon face in real-time. This paper focused on the issue of video face cartoonlization. Further more, we researched the face exaggerated methods based on average face. The detailed research content is as follows:1. We proposed an improved interactive face cartoon method, which can synthesize better cartoon face in complex background with simple user interaction. Firstly,we proposed a face cartoon method based on ERT model. Besides, to improve the synthetic effects of outline in cartoon face, we used the quadric b-spline interpolation method to amend. Finally, to improve the segmentation results of hair region in complex background, we used open operation in two blocks.2. We proposed a method to synthesize cartoon face video, including similar frame lookup based on the state of organ and cartoon video synthesize based on ERT model. At this stage, to improve the accuracy in similar frame lookup, we took the organ state as criterion which judges the similarity between two frames. In cartoon video synthesize based on ERT model part, it decomposed the organ motion into rigid motion and flexible motion firstly, then compressing the video to improved the feature point detection speed, and drive the cartoon face according to the detected motion parameters.3. We proposed two method to get exaggerated face based on average face, including face overall exaggeration and partial exaggeration for the most different organ. The former constructed the face feature vectors both in average face and tested face, and exaggerated all the difference between them. The latter judged the most different organ by the Euclidean distance and the type of face, and then only the most different organ and face shape will be exaggerated and distorted.4. A system of face cartoon video was given to complete all the algorithms. The system can implement the function of face video cartoonlization and face exaggeration by simple interaction.