Crop Image Mosaic Algorithm Based On Sift Feature Points

The feature-based image mosaic technology has become the main research direction in image stitching technology because it can make full use of the information in the image and make the registration between images more accurate.In the selection of the type of feature,the SIFT(Scale-invariant feature transform)algorithm is the main feature type in many studies because of its unique uniqueness,high quantity,strong robustness and good scalability.In this paper,the problem of extracting too many feature points in the non-overlapping region and slow processing of high-resolution crop images when using traditional SIFT algorithm for crop image mosaic.This paper presents a new improved algorithm strategy from the local and the whole of the process of stitching.The main research contents are as follows.(1)The SIFT feature points are detected from the six crop images,each feature is matched by the BBF algorithm based on the k-d tree,and establish the matching relationship between the images.Then,the RANSAC algorithm is used to select the matching feature points,calculating the transformation matrix between each pair of images,the image mosaic between each pair of crop image is realized by image fusion.Finally,the first two groups of stitching results with the last group of stitching to complete the final 6 crop image stitching.Through the programming realization and the experimental test,aiming at the obvious contrast between the main scene images in the captured crop image,a fast feature point matching method suitable for crop image is proposed.This method realizes the feature extraction of the main scene in the image by improving the threshold of the pixel contrast in the image,and reduces the extraction of the useless feature points in the image.Experiments show that the proposed method improves the matching accuracy by about 10% compared with the traditional SIFT algorithm,and the speed is more than doubled when dealing with crop images.(2)Two problems exist in the process of crop image mosaic based on traditional SIFT feature points:(a)The feature points of the non-overlapping regions in the image do not participate in the calculation of the final image mosaic,which is very prominent for high-resolution crop images,adding a lot of unnecessary calculations and taking up a lot of memory.(b)Crop images have a lot of the repeated scenes,it is easy in the feature pointmatching stage to form the wrong match point pairs,increase the final transformation matrix calculation time.In this paper,we propose a stitching strategy to deal with high-resolution crop images.The strategy reduces the resolution of the image,obtains the approximate overlapping area between the two stitching images,and then gradually refines the overlapping regions.It realizes the detection of feature points only on the area of the overlapping of high-resolution crop images,avoiding the above two problems.It not only improves the speed of high-resolution crop image stitching,but also improves the accuracy of image registration.Experiments show that compared with the traditional SIFT algorithm,the matching accuracy is improved by 20%.