Section: New Results

Correlation and variational approaches for motion and diffusion estimation in fluorescence imaging

Participants : Denis Fortun, Charles Kervrann.

Paper under review.

In this work, we have compared a correlation-based approach and a variational method for both motion and diffusion estimation in representative cell biology studies in fluorescence imaging. The so-called Spatio-Temporal Image Correlation Spectroscopy (STICS) is widely used in fluorescence imaging to recover physical parameters (e.g. direction of flow or Brownian motion of molecules). We have investigated recent advances in variational dense motion estimation and we have proposed to adapt the variational framework to the estimation of diffusion (i.e. Brownian motion). We have demonstrated the influence of the regularization parameter in the variational approach and its ability to capture motion of individual or clusters of moving objects. We have evaluated the advantages and limits of the two approaches for different biological studies (see Fig. 8 ).

Partners:: Perrine Paul-Gilloteaux, Francois Waharte and Chen Chen (UMR 144 CNRS PICT IBiSA Institut Curie)

Figure 8. Analysis of STICS and variational methods on artificial image time series with three phases. First row: first frame of the sequence and temporal description of the 3 phases: F/Flux (i.e. directed flow), D/Diffusion (20 - 30 - 40 images) (left); coding maps of vector fields (middle and right). Second row: STICS analysis for each phase. The arrows show the direction of the displacement and the color code is used to represent orientation and magnitude of estimated velocities. Third and fouth rows: Variational estimation for image pairs of each phase with a low regularization parameter (third row) and a high regularization parameter (fourth row).

motion scenario color code map vector code map             STICS-based estimation of different flux and diffusion phases             Flux 1 Diffusion Flux 2       Variational method with two different regularization parameters             Flux 1 Diffusion Flux 2         Flux 1 Diffusion Flux 2