Image Matching Algorithm Based On Feature Points Of Research And Implementation

Image matching is aligning two or more images by finding their transformation andcorrespondence, which may be taken under the different shooting conditions. In someapplications, it can also be defined as: alignment and matching process between a targetpicture and a template picture. Image matching is key in the field of computer visionsuch as object tracking,3D reconstruction, OCR, face recognition, medical imageprocessing, weather image processing and so on. The basic way is to match the featureobtained from image, which contains significant information of the images and theprocess is stable to the rotation, gray change, non-rigid transformation, noise and otherfactors. On the basis of analyzing and summarizing image matching algorithm based onfeature points, this dissertation presents an improved algorithm based on SC algorithm.Firstly, on the premise of keeping a high reliability of the algorithm, number ofmatching points is reducing by improvement of feature point detection algorithm andthe use of clustering algorithm. Then, images are matching with the use of TPSalgorithm. At last, the algorithm proposed in this paper and SC algorithm isexperimented in the case of noise disturbance, rotation changes, non-rigidtransformation and so on. The results show that this algorithm is better than the SCalgorithm in the matching speed and matching accuracy.