When acquiring panoramic images in Scanning Electron Microscopy (SEM), image tiles are affected by distortions emerging from two different parts of a microscope. First, the specimen stage introduces random image translations, as a result of imprecise shifts between the neighbouring positions. Second, a global deformation arises in the microscope imaging system. This deformation is dominated by rotation, non-orthogonality and scaling of the image. Therefore, there are visible imperfections at the seams of the panoramic image. There are various methods to correct these problems while stitching; however, the final correction usually does not correspond to the real deformation.

The aim of this work is to become familiar with principles of forming panoramic images in SEM, and suggest a suitable mathematical model of the imaging process, which after establishing the correspondences in the tile overlaps allows estimating the unknown parameters of tiles deformations. While searching for the optimal parameters, the knowledge of the limits of the acquisition device can be used. The proposed model will be practically evaluated and its behaviour demonstrated by stitching a more precise panorama.

The text of the thesis will contain at least the description of the panoramic image formation process in SEM, its mathematical formulation and the proposed solution, including its evaluation on the real images, and the results discussion. The practical outcome of the thesis will be the implementation of the proposed solution in a suitable programming language.