J. Biedermann1
1Technische Universität Cottbus, Lehrstuhl Numerische und Angewandte Mathematik
For real-time color image processing tasks like quality checking of colored objects, very fast and efficient algorithms for preprocessing image segmentation, analysis and comparision with prototype images are essential.
Especially efficient and noise tolerant segmentation procedures which should be able to run parallel play an important part to reduce the complexity of the following image analysis.
Classical algorithms like region growing strategies as well as the widespread split-and-merge mostly fail in real time applications for several reasons. Therefore new approaches like hierarchical segmentation and color histogramm methods are discussed and compared for their time complexity and their applicability to real time processing.