Résumé
The detection of proteins and the classification of their temporal behaviors in live cell fluorescence microscopy are of utmost importance to understand cell mechanisms. In this paper, we aim at locating and recognizing temporal events in TIRF microscopy image sequences related to membrane dynamics. After segmenting the time-varying vesicles in the image, we exploit space-time information extracted from three successive images only to model, locate and recognize the two dynamic configurations of interest: translational motion or local fluorescence diffusion. A likelihood ratio test is defined to solve this issue. Results on synthetic and real TIRF sequences demonstrate the accuracy and efficiency of the proposed method.