A Simple and Robust Cell Detection Algorithm
Presented by Mr. Michal SMíšEK
Type: Oral presentation
Track: Mathematical Methods in Image Processing
In this talk we present our advancements in designing a robust but simple algorithm for detection of cells in biological image data. Simplicity of this algorithm is to be understood as reduction of its user-defined parameters, which results in reduced calibration time. A starting point for us was the FBLSCD (Flux-Based Level-Set Center Detection) algorithm, and we studied the impact of its parameters, namely curvature and advection coefficients, image intensity threshold and stopping time, on its behaviour. Consequently, we have developed the LSOpen (Level-Set morphological Opening) algorithm, which doesn't use curvature term and the stopping time is chosen automatically. This novel algorithm performs at least as good as FBLSCD, with reduced user input and calibration, thus making the process of cell detection more automatic.