A novel digital watermarking scheme based on the kinoform and discrete wavelet transform (DWT) is proposed. The watermark image to be hidden is first encoded into a kinoform by utilizing the iterative phase retrieval algorithm in a virtual lens system. The weighted kinoform is embedded into the low-frequency DWT coefficients (LL3) of the host image to accomplish the watermark embedding process. In the process of watermark extraction, the embedded kinoform can be extracted by subtracting the corresponding DWT coefficients of the host image, from that of the watermarked image. Then, the original hidden image can be retrieved by taking the real amplitude of Fourier transform (FT) information for the extracted kinoform. The robustness and imperceptibility of watermarked image with the different weighting factors are demonstrated and analyzed in detail. The feasibility of proposed approach have been verified by computer simulations.