To enhance the capacity, invisibility, and security of image steganography, an image steganography model named CryptoStegoNet based on holographic encryption and dense residual networks is proposed. The model first converts the secret information into a QR code, then processes it with holographic encryption technology, and embeds it into the cover image to generate a high-quality steganographic image. The secret information can be extracted in the reverse process. The DenseResidualGenerator module, which consists of skip connections, DenseBlock, and DenseResBlock, is a key component of this model. Additionally, by introducing the FID (Fréchet inception distance) loss, the loss function is optimized to better guide the network training, making the generated images visually and statistically closer to the cover images. Experimental results demonstrate that compared with other state-of-the-art steganography methods, the model achieves significant improvements in visual quality, steganographic performance, and security.