Preprocessing Analysis on Medical Image Retrieval Using One-to-one Matching of SURF Keypoints

The development of medical image technology is increasing very rapidly. Moreover, medical image growth is predicted to grow exponentially over the past decade. Much research on medical images has been done, one of which is medical image retrieval using colors, textures, and keypoints matching. Generally, the keypoints matching approach is more reliable against spatial rotation. However, the medical images mostly contain blur objects and noise, which affect the performance of keypoints detection. This study proposes a preprocessing analysis on medical image retrieval using one-one matching of SURF keypoints. The image retrieval performance with and without preprocessing is compared on the IRMA 2009 dataset. This analysis proves that the use of preprocessing can improve the precision by 17.78%, 17.59%, and 17.92% on top-5, top-10, and top-15, respectively.