Image registration using perspective transformation for multimodal analysis of crystalline defects in semiconductor wafers

Abstract Detailed characterization of crystalline defects is crucial for enhancing the performance and reliability of semiconductor devices, especially in wide-bandgap semiconductor wafers. Combining multiple characterization techniques provides a more fruitful and robust evaluation of defects than any single technique alone. However, images obtained with these techniques often exhibit positional deviations and geometric distortions due to differences in imaging conditions and methods. Therefore, precise image registration is essential for practical multimodal analysis. To correct those misalignments, we propose a two-step image registration method. It uses coarse alignment via a similarity transformation, followed by fine alignment via a perspective transformation. Similarity and perspective transformations are determined by the positions of two or four pairs of corresponding reference points, respectively. This method accurately aligns multimodal images. Our image registration method clarifies crystalline defects and enables reliable multimodal defect analysis in semiconductor wafers.