Single Image Super-Resolution Technique using Precision learning of Low Resolution Images

A single image super-resolution technique is used to generate a high-resolution image from a low-resolution image by combining Non-subsampled contourlet transform and example based learning method. Sets of filters are learnt at coarse and fine scale from training dataset to produce an image of higher quality. The proposed method shows improvement both quantitatively and qualitatively in generating images of higher quality. The PSNR values have improved by a mean value of 3.64 dB when compared with bicubic interpolation and a mean value of 1.42 dB when compared with a state of the art method.