In SAR images, direct geocoding accuracy is affected by the input parameters, such as orbit data, height values introduced in the ellipsoid equation, and approximate geodetic coordinates. Specifically, the approximate coordinates, required for every pixel in the image by direct geocoding, contain errors due to pixel misidentifications on the map, to topographic maps, DEM and gravimetric to orthometric model. This paper describes a method, which evaluates the accuracy of the approximate coordinates introduced in the direct geocoding method for any pixel in the image, without the 'true' geodetic values being known. The method is based on the 'pseudo-errors' produced by the subtraction of the calculated by the direct geocoding method coordinates, from the approximate input coordinates introduced in the method. Evaluation of the above method is performed using three test points and four reflectors. It is shown that for the pixels of the SAR image, the corresponding evaluated geodetic coordinates when introduced as control points in the indirect geocoding process may produce geocoding results of accuracy higher than 1.4 ERS pixels. The method is useful if no reflector is located in the study area, because it produces four times better accuracies, compared to those usually produced by indirect geocoding.