<p>This paper proposes an image encryption algorithm based on small-world coupled mapping lattice and cellular automata. First, the initial parameters are jointly derived using the key and plaintext to drive the CML to evolve under the red–black update, and the key stream and mask are adaptively generated by ECA and local entropy. Reversible mixing is achieved through scrambling and channel integer shearing + modulo 1 expansion. Subsequently, deterministic CFB diffusion and byte-level DNA substitution are performed to obtain the ciphertext. Decryption is precisely restored in the reverse process. Experiments on standard images show that this method achieves nearly uniform histograms, high pixel entropy, strong key sensitivity, and can resist common attacks such as noise and cropping. This framework is user-friendly, easy to parallelize and has strict security attributes.</p>