Leeds 13th Statistical Workshop on Shape recognition in 2-D and 3-D images

The workshop took place in Leeds, July 12th-14th, 1993 bringing together over 40 participants from around the world, including UK, USA, Canada, Spain and Israel. The theme of the conference was the study of shape and its applications in 2D and 3D imaging, with keynote speakers Prof. M. Miller of Washington University, St. Louis, USA and Prof. J. Kittler, University of Surrey, UK.

The participants were from a broad range of backgrounds and the content included a blend of theory and application. The range of applications was wide, and included recognition of biological shapes such as mitochondria and amoebae; modelling heart ventricles and prostate glands in medical imaging; recognition of handwritten postcodes and human faces; computer animation and computer aided design; testing for shape differences in skulls of monkeys; segmentation of human wrist X-rays; modelling movement of pigs in a pen; line detection in industrial inspection; the use of cross ratios for object recognition; fusing nuclear medicine and MR images; automatic classification of human chromosomes; and fusing two views using object motion.

Prof. Michael Miller opened the conference in animated style with a lecture entitled `Mitochondria, membranes, amoebae, neuroanatomies, track targets: studies in 1, 2, 3-D shape'. The main contents of the talk included global shape models, group transformations and distributions for Bayesian analysis. The primary focus was the mitochondria example, where the images contained cell-like blobs viewed under an electron micrograph in 2D. The approach involves the use of pattern theory for representing knowledge in complex systems - the subject of a recent read paper to the Royal Statistical Society by Grenander and Miller. His second talk involved a deep discussion of the jump-diffusions that are used for inference in complex scenes, where the number of objects in the scene is not known.

Prof. Kittler spoke about parametric shape detection based on statistical hypothesis testing and in particular the relationship with the Hough transform. There were 12 other interesting talks over the three days ranging from statistical shape analysis to the computer vision and engineering approaches to object recognition. Posters were also presented and discussed during coffee breaks.

The final morning session was sponsored by the COMIR (Centre of Medical Imaging Research, University of Leeds). Profs. Smith (Medical Physics), Mardia (Statistics), Kent (Statistics) and Hogg (Computer Studies) gave a wide programme of talks, with applications in medical imaging and computer vision. The conference was concluded with a practical session on a cluster of Silicon Graphics workstations, conducted by Prof. Colin Goodall (Penn State University). The competitive task involved identifying objects by using cross ratios, with Dr. Tim Cootes (Manchester) being awarded the prize.

This is the second year that the workshop has operated in this conference format at the impressive facilities of Fairbairn House, Leeds. The active participation of invited and contributing speakers has proved to be very fruitful.

On the leisure side a conference dinner was arranged for the first time this year. With after-dinner anecdotes from Mike Smith this proved to be enjoyable and successful. Success was also had by a group of conference participants, `chaired' by Michael Miller - we won the local pub quiz!

Dr Ian Dryden, Department of Statistics, University of Leeds, Leeds, LS2 9JT. U.K. e-mail: iand@amsta.leeds.uk