Representação paramétrica de animações cartoon utilizando NURBS

Abstract

Cartoon-style animations and images, such as celluloid-style art, are a growing medium of artistic expression. As a result, the demand for new related technologies also grows. For example, a lot of effort is required from animators when creating cartoon animations due to their creation being done frame by frame. There is a need for means for interpolating new frames from the originals in order to lessen this effort. This work aims to have as main contribution the creation of a parametric representation of cartoon animations using NURBS. This creation will be frame rate independent. In addition, most of the time a smaller amount of data will be required for its storage, due to the low complexity of the input information. This work aims to explore the vector property of these animations, due to their low complexity of colors and contour lines, so that, then, it is possible to assemble a three-dimensional object through the control points that represent it, using NURBS. In this case, the third dimension will be the temporal representation of this object in the animation progress, thus allowing the interpolation of new frames from the original ones. In this work we will assume that the control points that represent the object are already collected and are necessary for the creation of the representation. The results show us that it has the potential to become a viable alternative to deep neural networks, which until now have been used for the interpolation of new frames. Thus, it will become possible to smooth the incoming animation with several extra frames, and still maintain fidelity to the original shape of the object. Due to the need to receive only the control points that represent the object and its flags term from NURBS and frame for the creation of our representation, noting that the proposed representation reached 84.23% of data emotions in relation to the original video representation for the cases studied.