Construction of intermediate views for multi-viewpoint 3 DTV systems

Any future 3 DTV system which preserves parallax of motion for the viewer must provide a multiplicity of images (multi-viewpoint). Any way of generating such intermediate views from only two original images (L and R) would be economically attractive. Particularly convenient for data reduction would be the transmission of only one of these images, together with so-called disparity information on the differences between the two. The author compares two methods of generating the second original image from the first plus disparity information, one based on edge detection (Marr and Poggio) and one based on correlation. The advantages of the second method are discussed.