SEM resolution improvement using semi-blind restoration with hybrid L1-L2 regularization

Scanning electron microscopy (SEM) resolution is limited by many factors that include sample specific properties, microscope stability, noise, the three dimensional nature of the sample and the excitation volume, and the spatial distribution of electrons in the probe known as the point spread function (PSF). If all, but the latter are optimized, the loss of resolution is principally due to blurring by the convolution of the PSF with the structure of interest. Image resolution can then be increased by deconvolution combined with the mathematical process known as regularization. To accomplish this task, a novel high resolution semi-blind image restoration technique incorporating hybrid L1 and L2 regularization terms has been developed. The original optimization is divided into the efficient solution of three subproblems, and has been validated with a variety of actual SEM images.