A Preliminary Study On Image Mosaics Technology

The thesis is based on the Project supported by National Natural Science Foundation of China----Researches on the Key Technology of Large-scale Detection of Coordinate, which aims at detecting 2D/3D coordinate of large-scale objects (more than 1m in diameter) with the existing CCD. At present, the ordinary CCD can detect the objects in a small field with high-resolution. However, it is difficult to detect the objects more than 1m in diameter. If the detecting range of CCD is expanded, the lower resolution can't meet the requirements of the testing of the coordinate. The ordinary CCD can only get partial images of the objects. And the equipments, such as panoramic camera and wide-angle lens, with which we can get a panoramic picture of the objects are comparatively expensive, thus not widely used. To solve this kind of problem, in this thesis, we think that the detecting range of CCD can be expanded if we take as many photos as needed with single or more CCDs and put them together by taking advantage of the redundant information between the images. But it is difficult to meet the accuracy of the experimental requirements using common methods of image mosaics. Therefore, it is of great practical significance to research an accurate and fast algorithm for image mosaics. It will be a breakthrough in limitations of CCD hardware, with the practical prospect.Taking the technology of image mosaics and the research results as basis, this thesis focuses on the F-M (frequency domain-based) and one-dot and one-line algorithm (feature-based), analyzing and summarizing the characteristics, performance and defects, then, a new technique of image mosaics being offered. This method integrates the advantages of Fourier-Mellin and one-dot and one-line algorithm. Through experimental analysis and comparison, this method discussed in the thesis can supposedly improve the image stitching algorithm accuracy and robustness.As for the noise and distortion that affect the accuracy of image mosaics, we suggest that, the interference can be minimized by geometric correction and filter processing before splicing the images.In the whole process of image mosaics, in addition to improving the algorithm, the various methods of image fusion are introduced. After analyzing the inaccuracy of the spliced images, the author draws a conclusion that many practical problems can be solved with the splicing algorithm.