This paper proposes a novel steganographic method based on tree structure. This method tries to reduce the amplitude of modifying the cover object to minimize the distortion. In data hiding process, each secret digit is represented by a (2n+1)-ary notational system. We employ complete tree structure to carry the difference between cover pixels and secret digits, and secret digits are embedded into the cover pixels by using tree structure. An optimum node distribution with least distortion is found by performing adjusting procedure, in which the amplitude of modifying is dispersed in all cover pixels evenly. The embedding algorithm obtains the stego object by performing addition between cover object and adjusted tree object. We can prove that our method can produce a stego object with least distortion. Theoretical analysis and experimental results show that our method is very efficient for hiding the secret message due to its simplicity and has higher embedding efficiency than other algorithms when the embedding rate is same.