Fluorescence spectroscopy is an image correlation technique to analyze and characterize the molecular dynamics from a sequence of fluorescence images. Many image correlation techniques have been developed for different applications [1]. But in practice the use of these techniques is often limited to a manually selected region of analysis where it is assumed that the observed molecules have homogeneous and constant behavior over time. Due to the spatial and temporal complexity of biological objects, this assumption is frequently at fault. It is then necessary to propose a robust method for discriminating the different behaviors over time from experience, as well as identification of the type of dynamics (diffusion or flow). This paper presents an original system of automatic discrimination and identification of spatially and temporally uniform regions for the analysis of molecular different dynamics over time by calculating STICS (Spatio-Temporal Image Correlation Spectroscopy) at different time lags. An evaluation of system performance is presented on simulated images and images acquired by fluorescence microscopy on actin cytoskeleton.