Research On Image Matching Based On SIFT Algorithm

Image refers to a plurality of images by matching certain characteristics algorithm to find the process of the same name points.Important subjects of image processing is image matching technology, it is also the foundation of some other image analysis techniques, image matching is an important means of visual information field.Since the image matching is widely used in many disciplines, many researchers study it for continuous improvement,and image matching algorithm is more and more mature.The image matching require multiple steps to complete, due to the diversity of the matching algorithm, you can get a wide variety of matching results.Algorithm based on image feature is now scholars most often used which SIFT(Scale Invariant Feature Transform) is the most representative of the algorithm.The algorithm is constructed using the Gaussian scale space descriptor,The algorithm in rotation, angle of rotation, affine transformation, and interference demonstrated a superior robustness,it also has strong specificity, enormous information content, scalability and other advantages.This paper studies based on SIFT algorithm image matching method, this article needle conducted intensive research image feature extraction and matching.First, introduces two classical algorithms: SURF(Speeded Up Robust Feature) algorithm and SIFT algorithm, and introduced the basic idea and processes of the two algorithms.Then, proposed an improved algorithm of SIFT, using circle rotation invariance as a reference, use this feature to described feature points in a circular area.Finally, we describe the SIFT improvement ideas and processing,using an experiment with two algorithms,then analyzed and compared the results.