Image upscaling based convolutional neural network for better reconstruction quality
T. Sree Sharmila, L. Megalan Leo · 2016
Excellent Resolution(ER) is kinds of technical that enhance the resolve of an imaging system, while improving image quality and the spatial resolution of the image. The image resolution is converted into low or high representation using upscaling factor. The sparse technic used to generate individual excellent resolution image. The complete image pixels should be good combination of spare elements with fixed lexical. Exceptional the Spare signal with combination of coefficient to produce the excellent output images. A sparse lexical (SL) pair containing low resolution and high resolution patches pairs and then using the representation of each low resolution patch via the low resolution lexical to predict the corresponding high resolution patch. It has faster lexical training than existing once and at the same time has better PSNR as well. It achieves significantly better PSNR than bicubic interpolation.