Home | Legals | Sitemap | KIT

Optical flow computation

Optical flow computation

Alexey Ershov

Optical flow computation

Figure 1: Vector representation of the estimated
motion field of thecockroach Periplaneta americana
during feeding.

The direct visual examination of image data is a non convenient and very time-consuming task. Moreover, a manual analysis is restricted only to small image sequences, while a typical radiographic sequence can consist of several hundred images. Additionally, only a small part of it is relevant from a scientific point of view. The experimental data have to be processed and analyzed carefully in order to be successfully interpreted and to obtain useful knowledge. In order to assist scientists with data manipulation routines and provide data analysis tools, a dedicated software package was developed. This software is based on the so called optical flow methods, which are among the key problems in the field of Computer Vision. Such methods solve the correspondence problem between subsequent frames of time-varying series of images. The incorporated methods are suited for a variety of scientific applications, such as motion analysis in living organisms (see Figure 1), pore formation in cellular materials, structural evolution, particles coarsening and movement, crack propagation, strain analysis, flow analysis in liquid materials and diffusion processes. By representing differences between images via a dedicated color-map, changes can be observed more easily and quantitative information like speed and direction of motion can be derived. Time-resolved data analysis software shows very promising performance for all kinds of dynamical X-ray data and is under intense further development.