Aiming at the problem of image superresolution reconstruction based on wavelet transform, an image reconstruction algorithm combining wavelet denoising and generalized regression neural network is proposed. First, the lowfrequency and highfrequency parts of the image are decomposed by an integer wavelet transform. Then the median edge detection is used as a predictor and an error map is set prior to encoding. Finally, the principle of the traditional generalized regression neural network is analyzed, and the corresponding generalized regression neural network is designed. In addition, the maximum value of the expected value algorithm was used to optimize the parameters of the generalized regression neural network model. The validity of the proposed algorithm was conducted by the superresolution image reconstruction simulation experiment. Experimental results show that the proposed algorithm has good denoising ability and high reconstruction accuracy.