Use of an adaptive 3D-DCT scheme for coding multiview stereo images

Our work presents a novel way for compressing image sets that form multiview stereo images used in three dimensional display devices. These image sets exhibit high inter-image correlation that remains unexploited when classic image compression techniques are applied. Our efforts are focused in the development of a compression algorithm based on the three dimensional discrete cosine transform (3D-DCT) which is an extension of the two dimensional case, used widely in image compression. An adaptive 3D-DCT scheme is implemented based on the exploitation of the increased spatial redundancy of certain parts of the image set. This is primarily achieved by locating image parts that belong to the distant background and hence remain unaltered throughout the image set. This technique achieves high compression ratios while maintaining the quality needed for correct stereoscopic viewing.