The goal of the project Tracking fluorescence within cells is detection and tracking fluorescence-labelled organelles in video, recorded by fluorescence microscopy method. Those fluorescent organelles or vesicles are visualized expression of modified genes.
The problem in detecting vesicles lies in the fact that fluorescence images contain a lot of noise. What helps us in their detection is their shape , which is mostly circular, and the size that appear to be very similar within a given series of pictures. As average intensity of the vesicles can widely vary over the image , it is natural to use local analysing operators.
In this project, vesicles are characterized with their moments.
Moment images are computed in multiresolution with B-spline window functions as analysing functions. Although only à trous version of algorithm for computing moment images is used , as more appropriate for signal analysis, we implemented Mallat like algorithm as well.
Moment images at coarser scales are efficiently computed using two-scale property of B- splines.
Tracking process is carried out as template matching procedure.
In first approach the procedure is initialized with detected image of the first frame, and template matching is used to find the best match for every spot in the next frame. As similarity measure is used minimum l2 difference between grey values from subsequent frames.
In second approach, every frame of the video is detected, and spots from subsequent frames are shaped-matched.
The result of the project is a module that provides a visualization of tracking process for selected vesicle and outputs its graph of distance.