A New Image Restoration Technique for SEM

This paper proposes a new fast and effective method of image restoration to improve the resolution of SEM images. In our approach, image resolution is improved by deconvolution with the point spread function modeled as the intensity distribution of the electron beam at the specimen's surface. The beam intensity distribution under each imaging condition is estimated by electro-optical simulation to achieve high resolution. We propose an iterative technique for the deconvolution process with a cost function where the restored image at each iteration can be compared with the original image more directly. A wavelet shrinkage denoising algorithm was applied to efficiently suppress noise amplification in noisy images, The empirical results demonstrate the outstanding improvements in both resolution and noise suppression. The proposed iterative method also speeds up deconvolution by about 3 to 50 times more than the conventional Richardson-Lucy method.