Research On The Techniques Of Digital Image Matching

Image matching technology is the extension of human visual cognition, and it is an important area of artificial intelligence. With the development of computer technology, Image matching technology more and more penetrates into our daily lives, gaining a wide range of applications. Due to the invariance of scale, rotation, illumination, SIFT (scale invariant feature transform) descriptor is commonly used in image matching, and becomes a new hot spot of image matching technology research. This dissertation focuses on the research in image matching technology, and the main content is described as follows:First of all, digital image processing and its applications are introduced, including the background, the significance and its historical status. Besided this, some basic knowledge of the image matching technology are discussed.Secondly,the basic idea of scale space theory is introduced. Besides, the construction of Gaussian scale-space is also discussed. Focuses on the construction of Gaussian scale-space and image pyramid structure and on this basis, multi-scale representation of image is obtained.Thirdly, the image matching technology is introduced in detail, at the same time, Scale Invariant Feature Transform (SIFT) descriptor is discussed. SIFT indicates every feature point using feature vectors with 128 dimension. But high dimension affects the speed of image matching. For the above mentioned issues, in the fourth chapter, dimensionality reduction in high-dimensional dataset is introduced. In addition, several key methods of high dimension dataset reduction are introduced in detail. Moreover, interpolation and its applications of image processing are discussed.Finally, in this dissertation propose a new method using image interpolation techniques to reduce the dimensions of feature vectors. The experimental results verify the validity of the method.