With about 120 islands located on 5 million square kilometers in the South East Pacific, French Polynesian islands are sometime difficult to reach and study precisely the vegetation. Remote Sensing is a very promising tool for vegetation mapping of these islands. Searching for a standard vegetation index, we found that the vegetation indexes based on low-resolution sensors (NDVI, RVI, etc…) are not precise enough to map the vegetation because of the small surface of most islands (a few square kilometers). We therefore applied those vegetation indexes formula on medium resolution images (MASTER) and higher resolution images (Quickbird) but the capability of these vegetation indexes to discern various classes of vegetation is limited. Two examples on Moorea (Society Archipelago) and Tubuai (Australes Archipelago) are discussed. We thus propose SARvi: a new empirical and simple vegetation index based on AirSAR Data, using the vegetation polarimetric properties. A study of this index shows its capability to distinguish more details in the vegetation structural properties than the previously studied vegetation indexes.