High-resolution wafer transmission electron microscopy (TEM) images are essential for nano-level analysis in semiconductor manufacturing. Although super-resolution offers significant potential for obtaining high-resolution images, training super-resolution models for wafer TEM images remains challenging because of wafer TEM-specific noise, the lack of paired low- and high-resolution images, and the need for clear edge preservation. These challenges have constrained research on wafer TEM image super-resolution. To address these limitations, this study introduces a unified three-stage super-resolution framework specifically designed for wafer TEM images, consisting of practical degradation, image enhancement, and super-resolution modeling. Using this framework, we conduct a comprehensive evaluation of previously proposed modules at each stage and identify the best-performing combination tailored for wafer TEM image super-resolution. The newly discovered combination from our analysis has not been explored before and outperforms existing approaches by effectively removing wafer TEM-specific noise and restoring visually clear edges. Furthermore, we provide a detailed module-wise analysis to evaluate the contribution of each stage to the overall performance of super-resolution modeling. Our findings highlight the significance in addressing wafer TEM-specific challenges and provide a strong foundation for advancing super-resolution modeling of wafer TEM images.
Kim et al. (Mon,) studied this question.