This paper discusses open problems andfuture research regarding the recognition and rep-resentation of structures in sequences of either 2Dimages or 3D data. All presented concepts aim atimproving the recognition of structure in data (espe-cially by decreasing the influence of noise) and atextending the representational power of known de-scriptors (within the scope of this paper graphs andskeletons). For the recognition of structure criticalpoints of a shape may be computed. We present anapproach to derive such critical points based on acombination of skeletons and local features along askeleton. We further consider classes of data (forexample a temporal sequence of images of an ob-ject), instead of a single data sample only. This socalled co-analysis reduces the sensitivity of analysisto noise in the data. Moreover, a representative for awhole class can be provided. Temporal sequencesmay not only be used as a class of data in a co-analysis process - focusing on the temporal aspectand changes of the data over time an analysis of thesechanges is needed. For this purpose we explore thepossibility to analyse a shape over time and to derivea spatio-temporal representation. To extend the rep-resentational power of skeletons we further presentan extension to skeletons using model fitting.


Janusch, I., & Kropatsch, W. (2015). Novel concepts for recognition and representation of structure in spatio-temporal classes of images. In P. Wohlhart & V. Lepetit (Eds.), Proceedings of the 20th Computer Vision Winter Workshop Seggau, Austria (pp. 49–56). TU Graz. http://hdl.handle.net/20.500.12708/56171