Performance Assessment of Feature Detector-Descriptor Combination

Features detection and description among multiple images are widely used in many applications, e.g., feature matching, object categorization, 3D construction, image retrieval and object recognition. This paper evaluates combination performance of different feature detectors and descriptors. It will compare performance of detectors and descriptors combination on images under rotate, scale constraints and distortion such as illumination on different scene (bedroom, industrial and CALsuburb). An experimental result shows MinEigen detector has best result in number of detected key-points when handle rotate, scale and illumination and not affected with scene. SURF without external detector is the best when handle rotate and scale constraint in different levels and scene. FAST/SURF and Harris/FREAK are best combined against illumination distortion in different levels. This review introduces a brief introduction for providing a new research in feature detection field to find appropriate method according to their condition.