Super-resolution is a technique to improve the wide-ranging quality, sharpness of images, video frames, and multidimensional images with low resolution (LR) by increasing resolution. In this paper, we analyze three popular superresolution approaches, namely SRCNN, SRGAN, and autoencoders, trying to improve their performance when aliasing is present in the training data. In each of the deep learning models, internal improvements and changes were made to the structures to improve quality, and the evaluation of the metrics was carried out on three data sets. Among the tested algorithms, the best results were achieved on a deeply rebuilt autoencoder with implemented recursive layers while maintaining a similar number of parameters, as well as on an improved generative adversarial network (SRGAN) belonging to the family of generative ad hoc networks. The analysis of the experimental results allows us to conclude that deep super-resolution models can indeed obtain satisfactory effects when trained with proper anti-alisaing techniques.

Ultralow-power electronics will expand the technological capability of handheld and wireless devices by dramatically improving battery life and portability. In addition to innovative low-power design techniques, a complementary process technology is required to enable the highest performance devices possible while maintaining extremely low power consumption. Transistors optimized for subthreshold operation at 0.3 V may achieve a 97% reduction in switching energy compared to conventional transistors. The process technology described in this article takes advantage of the capacitance and performance benefits of thin-body silicon-on-insulator devices, combined with a workfunction engineered mid-gap metal gate.