With the rapid progress of information technology, ensuring the security of images has become a critical research focus. This study introduces a method that integrates chaotic image encryption with convolutional neural networks (CNNs) to enhance both security and efficiency. To create strong image encryption, the approach combines the sophisticated feature extraction capabilities of a CNN model with the randomness and nonlinear mapping of chaotic sequences. We start by outlining the basic principles of CNN and chaotic image encryption.  A convolutional neural network (CNN), a deep learning model characterized by weight sharing and a local perceptual field, effectively reconstructs high-level image features. Meanwhile, chaotic image encryption leverages chaotic sequence generation and nonlinear transformations to scramble pixel values, ensuring secure encryption.  These procedures consist of feature extraction, pixel value mapping, key management, and chaotic sequence production.  Through the use of CNN to extract high-level image characteristics and dissimilarity operations between the chaotic sequences and image pixel values, the approach achieves high-strength encryption.  Lastly, we test the approach experimentally by contrasting it with more conventional chaotic picture encryption techniques.  The experimental findings show that the picture encryption technique offers advantages in computing performance and encryption/decryption speed, along with significant enhancements in encryption quality and security.