A three-dimensional(3D) Discrete Cosine Transform(DCT) method based on hexagonal sampling was proposed to fit the arrangement of hexagonal structure of the human retinal cells and to take advantage of the correlation among each color component of the color images. The method completed the conversion between the traditional rectangular sampling and hexagonal sampling according their relationships. Then, it proposed 3D Hexagonal sampling DCT(3D HDCT) on the basis of existing HDCT and verified its energy concentration. Finally, the spatial positions and color components of the color images in the same model were established, and the different images were transformed with different sub-plot sizes in a whole way by proposed method respectively. Experimental results show that the proposed method increases the compression ratio about 51.1% and the Peak Sigal to Noise Ratio (PSNR) about 16.3% as compared with that traditional rectangular sampling method, respectively. The results decrease the correlation among color components of color images effectively, and demonstrate that hexagonal sampling applied to image coding can improve sampling rates and decrease coding rates.