Image super-resolution reconstruction method based on regularization content mode weight prediction

The invention discloses an image super-resolution reconstruction method based on regularization content mode weight prediction. The method comprises the following steps: S1, carrying out low-resolution feature extraction on an input image through a feature extraction network; s2, performing regularization content mode extraction on the low-resolution features by using a regularization content mode extraction network; s3, performing position mapping on each pixel point on the super-resolution image, and determining position scale information and a regularization content mode of each pixel point; s4, using the convolution kernel weight prediction network to generate a convolution kernel weight for each pixel point on the super-resolution image, wherein the convolution kernel weight is matched with the position scale information and the regularization content mode of the pixel point; and S5, reconstructing the low-resolution feature of the corresponding position by using the convolution kernel weight of each pix