Two instances of image super-resolution using sensor raster scanning and deconvolution

摘要

Super-resolution reconstructs a high-resolution image from one or more low-resolution recorded images. It uses a variety of hardware techniques, such as sensor raster scanning, or software techniques, such as interpolation, iterative back- projection, and the regularized methods, and it is useful in a variety of applications, such as video surveillance, medical diagnosis and remote sensing. Two applications based on the same technique are presented here in some detail: first, the hyperspectral characterization of a THz beam generated by a Time Domain Spectroscopy device and second, the resolution-enhanced photography. In both cases, images recorded by the sensor during a subpixel raster scan were interwoven in a high-resolution image, subsequently deconvoluted with a constant square kernel of the size of the sensor pixel. For hyperspectral characterization, regularization was necessary, whereas for resolution-enhanced photography, simple deconvolution sufficed. Magee's approach to super-resolution, which does not use convolution, was analyzed and rejected.