Aiming at the good performance in computer vision for depth learning, a multi-focus image fusion algorithm based on depth learning is proposed. Based on the existing AlexNet network model, the convolution kernel size and step size are improved. The focused image blocks and the defocused image block are classified by using the improved scoring mechanism of deep learning network. Then, the correction matrix is used to correct the misjudgment image blocks. The boundary zone of image focus and defocus is subdivided and repaired. Six pairs of multi-focus images are randomly selected to verify the effectiveness of the proposed method. The experimental results show that, compared with other algorithms, the fusion results obtained by this algorithm can preserve the original high-frequency information of the image, and achieve good performance on four evaluation indexes of mutual information, edge information retention, entropy and average gradient.