The main goal of this task is to introduce how user and processing can communicate to allow the development of semi‐automatic solutions. To achieve this goal, pertinent information has to be extracted from the 3D fragments, partial objects, potential matching … Furthermore, visual feed‐back as to be integrated to legibly provide this information to the user. To this end, adapted rendering styles have to be developed. Another goal of this task is the user‐completion of missing parts by sketch‐based drawings.
- IPARLA: shape analysis, styled rendering, and sketching
- ESTIA: expertise in meshless analysis, and tangible interactors for sketching
The first step of this task is the definition of what kind of information can be extracted from eroded fragments and partial objects. This analysis will be based on geometry processing and curvature analysis. In order to deal with inherent noise in scanned data and eroded
objects, and to allow user selection of pertinent features, the proposed approach has to also be based on multi-scale analysis. For the exploration of the objects, similar to the traditional use of lights to reveal some of the aging or human marks, view‐dependent processing and image‐based processing will be privileged. Thanks to nowadays power of graphics cards and to the view‐dependent approach, the resulting feature analysis could be run in real‐time.
Since we focus on semi‐automatic solutions, the final approach has to allow user‐selection of pertinent features. Due to the view‐dependent analysis, theses features will be selected for a given scale and a given viewpoint. Nevertheless, once selected, such a feature has to be preserved in the analysis when the camera is moving, the object is moving, and the display resolution is changing. This would be possible by tracking these features thanks to the underlying 3D representation of objects.
We can thus afterward rely on this descriptor for following applications. Note that these developments can be based on existing geometry processing in case when the full descriptor would not be finished. One goal is the development of new real‐time rendering styles that provide the user with the information contained in the descriptor thanks to visual cues. We will focus on non‐photorealistic styles on one side, since they have proved their efficiency in conveying visual information. On the other side, for realistic styles, we will also investigate how shape perception can be enhanced based on the behavior of the human visual system.
Another goal is the user‐completion of missing parts. This work could be seen as either a guidance for the user‐assisted re‐assembly (Task 2.2) or object de‐aging (Task 1.3): it provides the systems with intuitions of missing or eroded parts. This work will be based on sketching
approaches for the user‐definition or deformation of global shape, and on the shape description, provided either by classical geometry processing or by our new shape analysis, to preserve and extrapolate the pertinent features. Furthermore, 3D modeling by painting
can also be investigated when it comes to the propagation of local geometric details.