Research Of Image Registration Algorithm Based On Intrinsic Marker

Firstly, this paper analyses the signification, concept, common methods and translation models of image registration. According to the starting point of our study, we classify the algorithm into there types: extrinsic marker-based, intrinsic marker-based and non image-based. The main task of our study is the research of image registration algorithm based on intrinsic marker which also can be classified into two types: low-level registration and high-level registration. Several problems discussed in this paper are listed as follows:1. The basic principle of low-level method is described. By making some analyses about image registration method based on low-level, achieved two images registration.2. The basic principle of high-level method is described. The critical technology of this method is feature extraction. We put emphasis on the different features, introduce what progress each field gains and point out what advantages and what disadvantages they have.3. Two comer detections is discussed. Many experiments are implemented on both Harris corner detector and SUSAN corner detector. We evaluate the performance of the two detectors from corner's effectivity, position accuracy, anti-noise and calculating speed etc. Experiments have shown that Harris corner detector can anti noise strongly and does not have misdetection. This Algorithm is efficient, rapid, and stable. Correspondingly, SUSAN corner detector is very accurate but it is very sensitive to noise.