In order to improve the robustness and invisibility of the image hiding algorithm based on wavelet transform, an image hiding algorithm is proposed based on the combination of Cellular Neural Network (CNN) and wavelet transform. First, the high bit planes of the secret image are encrypted by CNN encryption, and then the encrypted bit planes are hided into the singular value matrices of the wavelet sub band coefficients of the carrier image, realizing the encryption of the embedded data. The bit plane compensation method proposed by this article is used for information extraction in the information receiver. Experimental results show that this image hiding algorithm can still achieve good robustness and invisibility under the condition that the hiding data get a fifty percent reduction.