S. V. Kuleshov

Because of the active growth of 3D video format for entertaining purposes the interest in three-dimensional dynamic data representation increase. The existing technologies, for example DLP 3D represent only agreement concerning the displaying devices and the methods of 3D image formation based on two parallel video streams taken with standard for human stereo base (60 mm approximately) under definite and constant angle of view and viewpoint. Currently there is no existing format featuring compression of three-dimensional representation data suitable for TV broadcasting and for storing and transmitting of 3D content. The term voxel is used for 3D data transmission which stands for element of 3D image containing value on regular grid in three-dimensional space (analog of two-dimensional pixel). Voxel representations are often used for visualization and analysis of medical, scientific and technical information. Unlike existing data formats, the proposed True3D Vision universal format transmits unformed raster (not rendered) scene thus providing to change angle of view and stereo base values under playback. One of the first works using the idea of «motion as orientation» for video sequence representation was oriented on motion detection. On this basis an algorithm for video data compression and codec implementation was proposed. Spreading this idea on four-dimensional case allows obtaining the method to represent 3D data varying in time. Four dimensional form of video data representation (hyper cube) allows maintaining of local features as for one frame also as between frames and use this information for the compression purposes (redundancy). The efficiency of the considered True3D algorithm for time-variable 3D data is based on the fact, that after discrete cosines transform (DCT) the low frequency components are located compactly on hyper-cube element. After quantization and zigzag scan stages, obtained values are being coded with consideration of nulls repetition related to high frequency components of video data. As a result, only on quantization stage compression ratio reaches 90% with insignificant loss of visual quality