The invention discloses a single-image super-resolution reconstruction method. Mutual corresponding high and low resolution images are used as a training data set, an alternating direction method is used to train a high resolution dictionary and a low resolution dictionary from a training database, and then an inputted low resolution image is divided to obtain the set of original image low resolution image blocks. The high resolution dictionary and the low resolution dictionary are used to convert the original image low resolution image blocks into corresponding high resolution image blocks, the high resolution image blocks are combined to form a high resolution image, and finally the high resolution image is reconstructed to obtain a super-resolution image. According to the method, under an alternating direction method theory frame, high and low resolution dictionary training models are converted into three sub problems with analytical solutions to carry out iterative solution, and the computational efficiency is improved significantly under the premise of ensuring convergence. By using the method, the rapid super-resolution reconstruction of a single image can be realized, and the method has an important significance for many application fields of image processing and display.