Institutt for informatikk seminar: Thursday 14 feb, kl. 14:15
Digital distance transforms and their use in shape analysis
Centre for Image Analysis, Swedish University of Agricultural Sciences, Uppsala
The most difficult part of image analysis is usually segmenting the image into object(s) and background. However, this problem can in many cases, especially in industrial applications, be solved by the imaging process, using an adequate set-up with proper lighting. Once the segmentation problem is solved, the task is usually to measure various properties in the image or decide a diagnosis (e.g., correct or defect). This analysis task very often involves various aspects of "shape".
A very useful tool for shape analysis is the so called distance transforms (DT). Unfortunately, most general books and courses on image analysis ignores this tool, and solves the related problems in much clumsier ways, with higher computation complexity, and lower accuracy than necessary. In my talk I will introduce the concept of DTs and show a number of analysis tasks where they are useful.