The Abstract of my Diploma Thesis:
“Character animation utilizing hierarchical applications of surface based free form deformations”.

Note: The diploma thesis is in german language, but I translated the abstract into english.

Although this technique is well known and every 3D artist is used to it for many years, skinning geometrical objects is still a very tedious and time consuming process. Increasing requirements in terms of quality in todays games and feature film productions lead to a new level of detail and photorealistic computer generated images. Despite of techniques like subdivision surfaces, fine details can only be shaped on high resolution models—increasing efforts on skinning. To skin objects (e.g. character models) in a more efficient manner, several methods have been introduced over the years. Juxtaposed to methods simulating skeletal, muscle and tissue movements to deform a model, more general approaches have evolved to a serious option. Finer levels of control and an unlimited field of model classes with no need to share any correlation with physical existing objects are the major advantages. This thesis analizes how appropiate these general geometric methods can animate character models. The possibilities of driving objects by skeletal riggs are introduced as well as the application of Free-Form Deformations (FFD) developed by Sederberg and Parry, its enhancements and other solutions e.g. based on Bezier or S-Spline curves or surfaces. Especially shortcomings like the ones inherently linked to FFDs are accounted. The latter technique is very efficient in deforming complex objects by simply transforming the control lattice surrounding the desired space, thus warping embedded objects. On the other hand a parallelopiped lattice as utilized in the original version does not bear any visual resemblance concerning the object’s shape. Although several approaches leverage the topology of the lattice to support arbitrary shaped ones, applying these to an object or portions, hardly immediately result in an intended deformation. Therefore creating a conceivable result requires a selective and/or hierarchical combination of FFDs with a lot of manual tweaking. Having these drawbacks in mind, this work introduces a new surface based deformation approach, which allows the user for smoothly driving a maskable high resolution model by a low resolution mesh—greatly increasing the speed of work. Especially problems occurring when using euclidean distances (e.g. unmeant deformation of spatial proximate limbs) are elegantly solved. A comprehensive binding facility features selective rebindings of critical regions on the high resolution mesh using appropriate parameter settings at any stage during the workflow. Additionaly a feature called pose-animation allows for creating highly realistically poses of a character model in critical poses important for later animation tasks. Added to the plugin, the poses are automatically interpolated while animating i.e. the character will resemble the hand crafted pose in animation if the control mesh is near the defined position of the pose. Nevertheless the artist can freely access pose animation parameters at any time and what is more, is able to modify existing poses thus tweaking the whole animation at once—in the simpliest way possible.