Optical characterization of biological tissues is of real interest to improve medical diagnosis and in particular in the detection of precancerous tissues. We propose a new noninvasive method allowing the estimation of the anisotropy factor. This method is based on the image analysis of the Q- element of Stokes vector backscattered from the turbid medium. These Q-element images show specific patterns depending on g. Therefore the use of Fourier Descriptors (FD) on simulated data, to discriminate the specific geometrical features of the Q-element, enabled us to determine a linear relation between the anisotropy factor and six FD. This method was applied on experimental data obtained with calibrated solutions. The anisotropy factor was estimated with a maximum relative error of 13 %.