The detection of an interface separating two zones of low contrast and evolving in a deformable medium is very challenging. The low-contrasted images describing a partially saturated granular medium are replaced by high contrasted images generated from the correlation error maps issued from a standard Digital Image Correlation (DIC) calculation. A robust algorithm has been developed to capture automatically the sought interface evolving in 2D space through time. Capable of detecting all morphological reliefs, this approach is completed by a sequence of steps to remove aberrant contours and track the main advancing interface. Consequently, adequate algorithms have been developed to determine physical and morphological properties that will give insights on the factors that monitor the interface propagation.