In this paper we use kymographs and computational image processing to convert 3-D video microscopy data of
intracellular motion into 1-D time series data for further analysis. Because standard tools exist for time series analysis,
this method allows us to produce robust quantitative results from otherwise visual data. The kymograph-based approach
has an additional advantage over standard particle-tracking and flow-based image quantification algorithms in that we
can average out camera noise over the spatial axis of the kymograph. The method has the disadvantage that it removes
all spatial information. For this reason we see this method as a complement to rather than a replacement of standard
The standard problem we are trying to address in our work is how fluorescent proteins in one cellular compartment are
injected into another cellular compartment. The proteins travel at constant speed along a fixed spatial path, so a 2-D
kymograph produced from a trace along this fixed path will tell us about the injection history into this second
Our algorithm works by first taking a Radon transform of the input 2-D kymograph. We next make synthetic
kymographs by backprojection. The angle with the best correlation between the original kymograph and the backprojection
determines the dominant speed of the moving particles as well as the angle of the 1-D projected time series.
Time series are then analyzed with standard tools to determine the peak size distribution, the peak interval distribution,
the autocorrelation and the power spectrum.