AISB opportunities Bulletin Item
PhD Scholarship in Robust Statistical Fitting for Computer Vision Applications, Swinburne, Australia
Swinburne University of Technology, Faculty of Engineering and Industrial Science, Contact: Assoc. Prof. Ali Bab-Hadiashar (firstname.lastname@example.org) The aim of this project is to study the essential (and ubiquitous) problem of automatically segmenting visual data into meaningful parts - a form of model fitting: but in the face of multiple objects, missing and noisy data, large data volumes, and unknown form (model) and number of objects. These latter characteristics lift the problem beyond the reach of standard statistical fitting approaches. There have been studies concentrating on optimal fitting (Kanatani 1996) (ignoring such things as outliers, multiple structures, data sample density etc.), or on robust fitting in the presence of multiple structures ((Meer 2004) not only provides a good overview if these approaches but also a good discussion of why traditional robust statistics is inadequate for such a setting) but generally ignoring small sample issues, model selection, and a plethora of other practical issues. This project is an ambitious attempt to provide a more complete theory and practical methodology. (Kanatani 1996) K. Kanatani, Statistical Optimization for Geometric Computation: Theory and Practice, Elsevier Science, Amsterdam, The Netherlands, 1996. (Meer 2004) P. Meer: Robust techniques for computer vision. Emerging Topics in Computer Vision, G. Medioni and S. B. Kang (Eds.), Prentice Hall, 107-190, 2004. The project will focus on range data (i.e., laser scan and 3D reconstruction from images) and motion estimation/segmentation. It will build upon our previous work in this area (see references below). A. Bab-Hadiashar, N. Gheissari. Range Image Segmentation Using Surface Selection Criterion, IEEE Transaction on Image Processing, 15(7), 2006-2018, 2006. N. Gheissari, A. Bab-Hadiashar, and D. Suter. Parametric model-based motion segmentation using surface selection criterion. Computer Vision and Image Understanding, 102(2):214-226, 2006. K. Schindler and D. Suter. Two-view multibody structure-and-motion with outliers through model selection. IEEE Trans. Pattern Analysis and Machine Intelligence, 28(6):983-995, 2006. P. Chen and D. Suter. An analysis of linear subspace approaches for computer vision and pattern recognition. International Journal of Computer Vision, 68(1):83-106, 2006. R. Hoseinnezhad, A. Bab-Hadiashar, and D. Suter. Finite sample bias of robust scale estimators in computer vision problems. In Lecture Notes in Computer Science, International Symposium on Visual Computing (ISVC06), volume 4291, pages 445-454, Heidelberg, 2006. Springer-Verlag. Candidates should have engineering or science background (with strong mathematical and statistical skills) and experience in image analysis and software development would be beneficial.