We propose a new image processing approach to construct intermediate parallax-images from parallax-images at coarsely sampled viewpoints for Holographic Stereograms. Compared with the usual method that requires a large number of parallax-images, this approach reduces recording labor. In our approach, the intermediate parallax-images are made by projections of 3-D surface data, which are computed from several given parallax-images. It uses a new technique to reconstruct 3-D surface data of object scenes. The 3-D surface data is reconstructed from original parallax-images and camera positions. Almost all conventional techniques of stereo-matching need preprocess of feature extraction, and match only one stereo-pair at once. This technique simultaneously treats all given parallax images to find the depth map. In addition, feature extraction is not required in the matching process. Therefore it is possible to apply to almost all kind of scenes for holographic stereograms.