In most cases, for reinforced thennoplastics, a knowledge of average filler length is not enough infonnation. The entire granulometric filler distribution needs to be known, if structure is to be correlated with properties. We have developed a technique of granulometric detennination based on image analysis to analyse rapidly the evolution of the granulometry of short fiberglass. Once images in grey levels from polarized light microscopy have been acquired, each selected fiber on the frame is individualized and labelled before measurement and data treatment. For the fiber, the size factor chosen is the length, or the maximum Feret diameter. With this tool, we have been able to study more precisely the evolution of the fiber length distribution during processing with a polypropy lene matrix. Finally, we elucidate the dependence of mechanical properties such as impact and tensile strengths on glass filler size.