Single Image Super-Resolution Based on Total Variation Regularization with Gaussian Noise

The single image super-resolution to improve the image quality the texture in magnified image has been proposed. Our method uses a given low-resolution image only to generate high-resolution image, it cannot use extra high-resolution images. The single image super-resolution can improve the edge of image quality in magnified image. However, improvement of texture of sharpness in magnified image is difficult to generate high frequency components using only given input image. We propose single image super-resolution to improve texture of sharpness in magnified image by additional components of Gaussian noise. We can success to generate new higher frequency components by using Gaussian noise. Moreover, we propose assessment measure of image quality for super-resolution. In our experimental results, texture in magnified image by our proposed method is improved compared with other super-resolution methods.