DIKU Bits: Registering the big picture in data science
Jon Sporring, Professor, Image Analysis, Computational Modelling and Geometry Section (IMAGE), DIKU.
Modern imaging devices such as microscopes and synchrotrons provide data which are often too big and too complex for humans to analyze efficiently, and often we must rely on computer programs to extract knowledge from data.
Developing programs to analyze images give us novel, objective, and reproducible insights into images, which for some tasks are difficult to match by humans. For other tasks, it is a real challenge to develop versatile programs and human interpretations are to be preferred.
Microscope images is a particular interest to me since an increasing number of different microscope techniques are being developed which produce views of structures at a subcellular level, at unprecedented resolution, and in increasing data quantities.
Further, since the images vary greatly in appearance and quality, techniques from computer vision and medical image processing are only partially applicable. Often improvements of existing or invention of new algorithms are required, which I find a great source of inspiration.
In this talk, I demonstrate how image analysis is used for statistical models of ultrastructures in nerve cells and I will give examples of how we have improved existing and developed new algorithms for image analysis.