Research On Feature - Based Image Mosaic Algorithm

The21st century is an era of rapid development, with the rise of remote sensing technology, aviation technology attracted a growing number of countries concerned, in this condition, image mosaic technology has been extensively studied. image mosaic technique is composed of a sequence of images of the many panoramic images, through this technology, Humans can better to understand the nature and explore nature, in this case, the image mosaic technology has been widely used. This paper studies the method of feature-based image mosaic, including the following aspects:First, in this paper, the basic theory and key technologies used by the images mosaic were studied, including the basic coordinate system, homogeneous coordinates, transformation model, image interpolation, image registration, image blending. In this paper, through the analysis of the methods used in the image mosaic, the feature-based image mosaic method has been universally recognized and attention by academics.Secondly,in this paper, Image mosaic method based on SIFT feature points were analyzed, including SIFT is a rotating, light, scale transformation remain invariant feature points, but because of its128-dimensional feature descriptors calculation volume is very large, Therefore, the need to reduce the amount of calculation. firstly, using phase correlation method can effectively get the overlapping area of the two images, then, extract the SIFT features in the overlapping areas of the image, final, image mosaic.Finally, the image mosaic algorithm based on Harris corner operator is studied, first, the feature point extraction, secondly, matching the extracted feature points, Then,RANSAC method except to outliers,finally, graph cut combined with pyramid blending algorithm for image blending, experiments show that this algorithm can achieve good image mosaic.