A Geometry-Distortion Resistant Image Detection System Based on Log-Polar Transform and Scale Invariant Feature Transform

This paper presents an image detection system based on Log-Polar Transform (LPT) and Scale Invariant Feature Transform (SIFT). Unlike other schemes that extract features from the original image, the presented scheme extracts features from the transformed image by LPT. Moreover, the presented scheme utilizes SIFT to extract geometric-invariant features from the LPT images to achieve greater robustness and resistance to geometric distortion. When given a suspect image, the scheme compares the extracted features from the host LPT image and the suspect LPT image to determine similarity. The experimental results show the presented scheme can achieve high recall and precision rates even when the duplicate image is modified and not exactly the same as the host one.