Abstract: A system encodes video frames in accordance with the MPEG standard, in real-time. Motion vectors are calculated during the forward prediction of B frames and P frames using a telescoping procedure. Similarly, a telescoping procedure is used for calculating motion vectors for B frames during backwards predictive coding. The use of telescoping for the backward predictive coding results in B frames being coded in the wrong order. The correct order for B frames is re-established by writing the frames to a buffer in the order in which they are generated and then reading the frame data from the buffer in the order in which they are required for transmission or recording purposes. The use of the telescoping procedure reduces significantly the amounts of computation required for calculating motion vectors during predictive coding for frames which are displaced by more than one frame period.

Abstract: Telescopic predictive coding of video frames reduces the number of pixels in a reference frame which are required to be processed. However, given that offsets for adjacent blocks may differ, it is difficult to parallel process for real-time coding. After first-frame motion vectors have been calculated, a region of the reference pixels is read which is large enough to accommodate the direct calculation of motion vectors for the second frame, thus allowing pixel values to be processed in parallel. However, although all of the read pixels are made available to each processor, the processors only latch the values which are actually required for the particular motion vector being calculated. This is achieved by supplying an enabling signal to the processors, which is determined by processing the first-frame motion vectors.

Abstract: A method of coding a video signal for transmission includes the steps of coding data representing the video signal by a base layer coding operation which includes base layer quantizer having a base layer quantization step size to provide coded video data for transmission; deriving inverse-coded video data by carrying out an inverse base layer coding operation on the coded video data; deriving difference data from the data representing the video signal and the inverse coded video data; and coding the difference data by an enhancement layer coding operation only when the energy of the difference data exceeds a variable threshold, the threshold being inversely proportional to the base layer quantization step size.

Abstract: A motion vector for motion-compensated prediction in an inter-frame differential video coder is derived by comparing each block in a row of a current frame with the corresponding region, and positionally shifted regions, of a previous frame. The blocks are processed row by row, and scanned in a vertical column-by-column scan with each of the shifted regions. Two or more rows may be processed at a time. For .+-.7 pixel search, two or four comparison processors are provided, one (or one pair) processing odd-numbered blocks and the other (or other pair) processing even-numbered blocks. For a .+-.15 pixel search, four or eight processors are provided. The vertical scan is provided using a serial FIFO store including eight 8 tap SIPO sections.

Abstract: A motion vector suitable for motion-compensated prediction in an inter-frame differential video coder is derived by comparison of each block of a current frame and with shifted corresponding region of a previous frame with the regions of the previous frame. Rather than dealing with each block in turn, the apparatus carries out all comparisons involving a line n of the video signal before commencing comparisons involving n+p (where p is the number of lines encompassed by a block).