For the past few years, the scale of integrated circuit chip technology feature has been narrowed down to tens of nanometers. Intrinsic wiring hierarchy has reached twelve layers, and the spacing between layers can reach a minimum length of thirty nanometers. It’s a great challenge to acquire chips with perfect surface of all circuit hierarchy once for the large scale integrated circuit reverse analysis. Instead, it usually needs several times of chips processing to compensate the defects mutually, such as surface contamination. So it is necessary to inpaint the spotted IC images for making the whole-chip image library in the nanometer level integrated circuit analysis. Due to the nanometer level sophisticated circuit structure, it is inapplicable to large defect area image inpainting issue with methods based on neighborhood information of the spotted area, such as image inpainting based on partial differential equations or image inpainting based on texture synthesis in the background of microscopic image mosaic. According to the characteristics of defect area in IC chips, we design a large defect area image inpainting approach based on thin-plate spline algorithm, after analyzing the difficulty of large defect area image inpainting issue in the background of image mosaic. Through locating the spotted area by manual method, we warp the additional acquisition images to repair the spotted images with TPS, which greatly increase the applicability of the algorithm. Additionally, several methods are discussed to solve the ill-conditioned problem that we encounter in above disposal, including the Least Squares method, the Regularization method and the Moore-Penrose method, which has draw a conclusion that the Moore-Penrose method can get the best stitching result. Since without the dependence on similar structures in the IC image, the method can apply to other image processing fields, such as spotted brain slices images inpainting. The method proposed in this dissertation has been applied to solve large defect area image inpainting issue in the background of microscopic image mosaic. And corresponding inpainting process has been designed and a practical software tool for defect microscopic image inpainting has been developed, which has been integrated into large scale electron microscope image mosaic software Mosaic and has been applied to several million gate level integrated circuit chips analysis project successfully and has solved the defect microscopi...
修改评论