Date:	Thu, October 22, 2009
Time:	15:15
Place:	Research II Lecture Hall

Abstract: Normal multi-scale transforms are a relatively new and promising tool for compression, filtering, geometric modeling, etc. The transforms are non-linear, due to the different type of the data involved (vector-valued coarse-level data, and scalar details) which makes them difficult to analyze. Moreover, there are very few theoretical results available. In this talk we will present the known theory, namely the case of normal MT for curves in 2D based on linear, local, dyadic prediction. Open questions and possible future research projects will be discussed at the end of the talk.