Real image super-resolution model and method based on corner guide cascade hourglass network structure learning

The invention discloses a real image super-resolution model and method based on corner guide cascaded hourglass network structure learning, and the model comprises: a multi-scale feature extraction unit which extracts the features of multi-scale information of an input image through employing a cascaded hourglass network structure; the regional reconstruction unit that is used for respectively reconstructing a plurality of initial super-resolution images by utilizing multi-scale features of different depths; the regional supervision unit that decouples the high-resolution image into flat, edgeand corner regions by using a corner detection algorithm, and supervises each initial super-resolution image respectively; an angular point guide reconstruction unit which uses the extracted information of each area of the image; and the gradient weighting constraint unit that is used for weighting the loss function based on the gradient information of the image and enhancing the fitting capability of the angular point r