Overlapping satellite images are one of the key solutions in both generating and updating of spatial data like digital elevation model (DEM). To do this, rational polynomial coefficients (RPCs), as an essential Meta data for satellite images, are regularly used. Nonetheless, these coefficients due to some systematic errors in their gathering lead to an erroneous DEM. Accordingly, in this study a new method is proposed to make it possible to obtain an accurate DEM using automated bias compensation of the RPCs in object space. In this method, RPC biases on the resulting DEM are nearly removed using a 2.5D matching procedure between IRS-P5 derived DEM and the national-cartographic-center DEM (NCC DEM). The experimental results of this study showed that using both 10-meter NCC DEM and stereo images of IRS-P5 satellite images can achieve about one pixel level of accuracy or better in both DEM generating and updating.