An Auto-rigging Method For 3D Cartoon Models

The auto-rigging method for 3D models is useful on many fields such as fast-modelling and animation design.The performance and generality are the focus of researches.The traditional auto-rigging algorithms usually use morphological or mathematical methods to extract the curve skeleton of the model and preserve the topology.However,the shape of the generated skeleton is totally determined by the input model which means the user cannot choose the skeleton type.Another problem is the large cost of the computing resource which cannot be applied on the mobile device.The auto-rigging method proposed in this paper has solved the mentioned problems.It can recognize the joints of the input cartoon model and a standard skeleton is embedded into the model.A neuron network is adopted to improve the quality and generality of our algorithm and a 3D-2D-3D transformation is used to achieve the high performance.This method can finish the auto-rigging procedure in 1 second on the mobile device.To improve the performance,this method change the input from 3D models to 2D outline images.Several candidate points are extracted from the outline image and the skeleton joints are recognized from these candidate points.The generated 2D skeleton then is transformed to 3D space and bound on the model.Since cartoon models need less accuracy than realistic models,the degree transformation results in very little quality loss with large performance improvement.Our method can generate different kinds of skeletons such as human skeleton,plant skeleton,and so on.The neuron network is used to classify the candidate points,which improves the generality of our method.Also,our method can accept different input such as 3D models or 2D images.A cartoon animation editor is provided in our work.The users can use this tool to easily design a short cartoon movie,which may be useful on the education and design regions.The contributions of this paper are in three aspects.Firstly,a 3D-2D-3D framework on auto-rigging is proposed to get a great improvement on the performance.Secondly,a skeleton extraction method based on 2D outline image is proposed in this paper.A neuron network is used in this method to enhance the quality and generality of the result.The method is easy to implement and has good performance.Thirdly,a useful tool of cartoon animation editing is implemented for the application of our work.This tool supports the users quickly generate cartoon models and edit short animations,and it is easy to use.