Polarimetric　SAR　interferometry　has　been　studied　widely　for　forest　observations.　This　technique　decomposes　the　local　scattering　centers　by　using　the　polarization　state　difference　among　them.　Therefore,　those　forest　parameters　can　be　estimated　precisely　with　this　method.　Recently,　many　methods　of　extracting　interferometric　phase　have　been　proposed,　among　which,　some　methods　that　are　developed　from　spatial　spectrum　estimation　theory　has　been　improved　gradually.　These　methods　always　use　MUSIC　or　TLS-ESPRIT　algorithm　to　estimate　interferometric　phase.　However,　those　methods　bring　some　confusion　of　the　inerferometrics　phases　from　different　scatter　centers　due　to　the　lack　of　scattering　centers＇　separation.　In　this　paper,　the　TLS-ESPRIT　algorithm　of　joint　estimation　of　inerferometric　phase　and　power　by　calculating　the　power　of　the　backscatter　signals　was　proposed.　Since　the　number　of　the　dominant　local　scattered　waves　and　their　interferometric　phases　in　each　image　patch　can　be　estimated　respectively,　the　proposed　algorithm　can　classify　the　phases　and　implement　the　separation　of　the　scattering　centers　by　utilizing　the　differences　among　the　powers　of　backward　scattering　signal　with　different　mechanism　in　the　same　resolving　unit.　Moreover,　in　this　method,　the　two　height　parameters　can　be　obtained　by　extracting　the　two　optimum　phase　parameters,　furthermore,　the　local　interferometric　phases　of　the　ground　and　canopy　can　be　detected　separately.　Especially,　this　method　has　an　advantage　of　performance　when　the　local　scattered　waves　are　uncorrelated　and　have　different　polarization　characteristics.　Therefore,　it　has　higher　accuracy　in　estimation.　In　this　paper,　the　principles　and　the　implementation　steps　of　this　method　are　addressed　in　details　and　its　effectiveness　is　tested　using　SIR-C/X-SAR　L　band　data.