Aiming at the problem that the most of image information hiding techniques need prior information and the difficulty of training, one information hiding method for images based on complex wavelet transform and adaptive pixels clustering is proposed. A new unsupervised population based optimization algorithm is utilized by the proposed method to cluster the pixels of secret image adaptively, then the support vector machine is adopted to select the optimal wavlet sub band of dual-tree complex wavelet transform of the carrier image, in order to maintain the imperceptibility of hidden information, the method treats the optimal wavlet sub band as carrier for hidden information. Experimental results show that the method realizes strong imperceptibility, and it not only has strong ability in resisting intrusions of steganalysis models, but also does not need any prior knowledge of the original carrier image and secret information during the information hiding process. It is a bind information hiding method.