An extension of our 2009 BMVC paper has been accepted for publication in the International Journal of Computer Vision. Besides a more detailed evaluation on the completenss of combinations of local feature detectors, it covers the theory for a new sparse keypoint detector based on the maximum entropy principle. The print version of the paper will appear in summer 2010.

The Springer proceedings of the 7th International Conference on Computer Vision Systems held 2009 in Liege, Belgium are published in google books. My own contribution "Evaluating the Suitability of Feature Detectors for Automatic Image Orientation Systems" can be found here.

The German edition of the Technology Review magazine has published an article covering industrial applications and potential of image processing and recognition systems, with many comments from Joachim Denzler who leads the research group in Jena. It is a short but interesting insight for interested non-specialists, focussing on the German market.
Read article

Aicon 3D systems introduces MoveInspect, a modular and flexible photogrammetry-based 3D coordinate measurement system for 3D inspection. Just before the introduction, our students Miriam and Birgit investigated the reliability and accuracy of such systems in a joint work with aicon. The work was part of their diploma thesis, which I supervised.

My oral presentation about the "Completeness of Coding with Image Features" at the British Machine Vision Conference 2009 in London has been recorded on video. It is now available at videolectures.net. If you're interested in the work, read the full paper.

I highly recommend listening to Alyosha Efros's invited talk, "What can the world tell us about an image?", and Andrew Fitzgibbon's motivating two-hour Tutorial on continuous optimization, covering parts of Triggs' excellent survey of bundle adjustment techniques.

We have set up a website for the SFOP keypoint detector that we presented on ICCV in Kyoto this year. The website includes a very slow implementation of the original algorithm. We are now working on a fast parallel implementation on the GPU. You find the paper describing the detector here.