The Research Of Image Registration That Apply To Visual Navigation

In computer assisted surgery, image registration provides the three-dimensionallocalization of navigation system. Image registration is the premise of three-dimensionalreconstruction. It can greatly improve the clinical diagnosis, monitoring, treatment andcurative effect evaluation of surgical operation level.Because the surface reflectance andnoise due to the images in the collection process, the image feature extraction results ofcurrent registration are often not very obvious.And many of the registration algorithm usesthe entire image pixel to calculate,thus cause a large amount of computing as well ascomputing speed reduce, so as to solve the problem of feature extraction ineffective andslow computing speed.This paper proposes the mark points and the image feature points to achieve theregistration.The main research work and research results are as follows:1、Design auxiliary point that is easy to identify and extract. Extract the marked pointregion, then calculated the center of the point. Propose to use the concept of chain code toachieve additional marker registration.Use the centroid chain code to do the first-orderdifferential in order to get differential chain code. By the the invariance of rotational andtranslational in differential chain code,get registration point;2、Analysis of SUSAN and Harris corner extraction algorithm, on the basis of these, putforward the improvement Harris corner detection algorithm. First,judge the intensity changeof the pixel and ignore the smaller part in order to narrow the range of detection. Then putforward the adaptive threshold method to extract feature points, and then use the idea of theSUSAN algorithm to select the extract characteristic points;3、Use normalized cross-correlation for initial registration with extracted corner, thenpropose using coordinates of feature points of similarity to the precise registration.The experiments show that the method of using external markers does not existmismatching points.The method is simple and can get precise registration points.The featurepoints extracted by improved Harris algorithm can be evenly distributed.After removing thefalse points it can get accurate registration point and after registration can get ideal effect.The registration methods that uses of external markers and image feature point for featureextraction has obvious effect, strong anti-interference ability, and less calculation, afterregistration can obtain the precise registration point.