Image matching is known as a vital process in digital photogrammetry. Despite development of many image matching algorithms, this process still has some difficulties in close range photogrammetry, due to geometric changes which are made by changes in viewpoint. In this paper, an effective and robust image matching approach is presented for wide-baseline image matching. In the proposed method, in order to perform matching operation, after extracting blob-like features in base and input image, using SIFT detector, an elliptical region is constructed for each feature . In the following, in order to control the created geometric changes resulted from changes in the imaging perspective, parameters of this ellipse are calculated by using second moment matrix. In addition, descriptor for each feature is constructed by normalization of the respective ellipse to a circular region with a constant radius. Finally, by applying nearest neighbor method, matching process is done and mismatches are eliminated by implementing epipolar constraint based on RANSAC method. Test results on different close range image datasets, beside the increased accuracy rate of 3 to 8 percent, represents significant function of the proposed method since the results are two times higher than the results finalized using the standard SIFT method.