SIFT; F-transform; image downscale; keypoints;

This paper analyzes the efficiency of the Scale Invariant Fea-ture Transform (SIFT) measured in terms of execution time and stabilityof detection for keypoint location. Our goal is to investigate the inu-ence of various downscaling methods which pre-process images beforethe SIFT is executed. For every downscaling method, we estimate thetotal execution time, the amount of detected keypoints and the successrate of keypoint matching. We show that the best combination with theSIFT is achieved by the F-transform downscaling algorithm.