Based on the DenseNet model, this paper explores methods for automatic image content recognition and typesetting. The traditional methods have some limitations in feature extraction and typesetting, so a new solution is investigated. In this paper, we improve the accuracy and efficiency of image content recognition by using the DenseNet model, and at the same time design intelligent image typesetting algorithms by combining with deep learning technology to realize the automated typesetting of images. The method is experimentally validated to achieve significant improvements in both image content recognition and typographic effects, reducing the need for human intervention and improving processing efficiency (Max: 88%) and quality. This study provides new ideas and methods for intelligent development in the field of image processing and design.