Improving Accuracy of Optical Flow of Heeger's Original Method on Biomedical Images

摘要

The paper is focused on the computation of optical flow from time-lapse 2D images acquired from fluorescence optical microscope. The Heeger's traditional established method based on spatio-temporal filtering is adopted and modified in order to solve issues that arose from this sort of image data. In particular, a scheme for effective and fast computations of complex Gabor convolutions is used. The filter tuning is changed to better support the detection of movement. The least squares fitting of the original method is also revised. A parametric study was conducted to assess optimal parameters. With optimal parameters, the proposed method showed lower average angular errors than the original. C++ implementation is available on the author's web pages.