Research On Endoscopic Image Mosaic Technology For Minimally Invasive Surgery

Endoscopic minimally invasive surgery is an important development in the field of surgery,endoscopy can be a good assistant to complete the clinical diagnosis and treatment,as well as minimally invasive surgery.Compared with ordinary surgery,minimally invasive surgery is through a tiny wound,by using endoscopic images information to observe its internal structure.Endoscopic imaging plays a very important role in the diagnosis and treatment of lesions.However,the characteristic of endoscopic imaging is that the imaging range is small and the scanning of larger organs can not be completed once.This affects the physicians' judgment in detail for the range of lesions and details,that isn't conducive to the qualitative diagnosis of lesions,image mosaic technology can be used to solve this problem.However,the problems such as the high mis-match rate for image registration,the real-time demanding,the image scale change and so on,this paper has carried on the detailed study of endoscopic image mosaic,the main contents are as follows :First of all,the basic process and key technologies of endoscopic image mosaic are systematically introduced,including the extraction of feature points,the generation of descriptors,the matching of key points,the calculation of transformation matrix and the image fusion,the feature points matching is the key.Secondly,the SIFT(Scale invariant feature transform)algorithm in the field of image mosaic has been studied in detail and improved.Mismatching occurs due to feature point matching,because of existing mismatch rejection algorithm is generally a single method,in order to solve this problem,we propose to combine the nearest neighbor matching method with the RANSAC(Random Sample Consensus)matching algorithm.That is,the distance between the nearest neighbor and the next nearest neighbor is used to determine the matching points,RANSAC algorithm is used to eliminate the mismatch points and find the homography matrix H of the corresponding frame.Then,the registration of the endoscopic images is performed according to the homography matrix.Finally,an in-out weighted fusion algorithm merges the registered images to eliminate the obvious gaps generated in the image stitching process and realize the panoramic endoscope images.Experimental results show that the stitching effect is good and the accuracy of feature point registration has been improved.Finally,the pyramid ORB(Oriented Brief)algorithm is proposed to meet the real-time requirements of endoscopic image mosaic.This method has more efficient feature extraction and representation than SIFT,which reduces the computational cost and memory requirements,improves the efficiency of image registration.The disadvantage of ORB is that it doesn't have the scale invariance,so the pyramid ORB algorithm proposed in this paper solves this problem well.The performance of pyramid ORB algorithm is compared with that of improved SIFT algorithm,the experimental results reveal that pyramid ORB algorithm has good robustness to changes in image brightness and noise,the accuracy is high,and the matching speed is about 10 times that of SIFT.