In this paper, a novel design of dual-tree complex wavelet transform based on least squares method is proposed. An effective design method based on allpass filter with FIR has proposed by Selesnick. However, this method holds two problems. First, optimal frequency characteristic cannot be realized due to the restriction of design parameters. Second, approximation of the half sample delay condition is restricted on low frequency. The proposed method can achieve the high design degrees of freedom by using least squares method. It is shown that the method can design more effective dual-tree complex wavelet transform than conventional one, in terms of sharpness of frequency characteristic and accuracy of half sample delay condition, directionality and shift-invariance, by simulations. In the end of this paper, the performance of DTCWT designed by the proposed method for practical application, image denoising, is demonstrated. The proposed method outperforms the conventional method.