Nov. 28 2008 10:00AM

Nov. 28 2008 11:00AM

Automatic analysis of 3D volume and shape data is an emerging filed in computer vision and pattern recognition. Especially the increasing use of new 3D imaging techniques, i.e. confocal laser scan microscopes, in biology and other live sciences create a demand for automatic 3D image analysis. In this talk, we present a generic approach which is suitable for a wide range of 3D image analysis problems such as 3D texture and shape description, object detection, classification and segmentation. This approach is based on the combination of local invariant features and learning techniques.

The talk will focus on the local features and will introduce several different methods for the extraction of rotation and gray-scale invariant features for multi-channel 3D volume data and 3D vector fields. Finally, we will present several applications of the generic approach on biological image analysis tasks and 3D shape retrieval.