| Image mosaic technology is an important research field of image processing. Image mosaic of power semiconductor device's images is a specific application in microelectronics area. In this paper, On the basis of analyzing the current mosaic technique and the special nature of the image , we introduced a proper method to mosaic the microscopic image. The goal of this paper is to mosaic the large number of microscopic image which collected under the microscope to a global territory panorama,and make some efforts and attempts to achieve automatic registration and stitching,and make some preparation for the future application of measurement and designs.The main work done in this thesis is as following:(1) Registration of microscopic territory image. Due to the particularity of the microscopic territory image, the general registration algorithm in spatial domain is difficult to achieve good matching results. This paper uses frequency domain phase correlation of the POC-based registration algorithm to registrate the image. Because of uneven illumination and the small overlapping region,Some images'registration is error. In this paper we do some effort to improve the original POC algorithm to reduce the mismatch. Through adding the Image Substraction and features block extraction to POC algorithm , effectively improve the registration accuracy.(2) A large number of microscopic images'global stitching. we research the currently major global stitching algorithm and analyze difficulties and problems in the method of eliminate error accumulation.In order to reduce the error accumulation in the global microscopic image mosaic, Based on the analyses of frame-to-mosaic stitching method,This paper using the idea of Merge sort algorithm to achieve the global optimal pyramid splicing strategy. Error accumulation is reduced down to pass, The experiment shows this method also reduces the overall complexity and computation.of frame-to-mosaic mosaic algorithm(3)Blending mosaic of. microscopic territory image. On the basis of analyzing the performance of the existing Seamless mosaic algorithm ,a improvement method based on the weighted smoothing method is proposed. By experiment shows that this method can basically meet the layout requirements of microscopic image of Seamless mosaic.In this paper, on the basis of existing image stitching techniques, in order to verify the effectiveness of the method we have carried out some experiments to achieve a microscopic image of the mosaic map, indicating the validity of the study. |