Pulse or digital communications – Bandwidth reduction or expansion – Television or motion video signal
Reexamination Certificate
1998-11-20
2001-10-30
Kelley, Chris (Department: 2713)
Pulse or digital communications
Bandwidth reduction or expansion
Television or motion video signal
C348S405100
Reexamination Certificate
active
06310915
ABSTRACT:
FIELD OF THE INVENTION
The present application pertains to video encoding. In particular, the present invention pertains to re-encoding a previously encoded video signal.
BACKGROUND OF THE INVENTION
The MPEG-2 standard is widely used for encoding video. According to this standard, pictures are both spatially and temporally encoded. Each picture is first divided into non-overlapping macroblocks, where each macroblock includes a 16×16 array of luminance samples and each block or array of 8×8 chrominance samples overlaid thereon. A decision is made to encode the macroblock as an inter macroblock, in which case the macroblock is both temporally and spatially encoded, or to encode the macroblock as an intra macroblock, in which case the macroblock is only spatially encoded. A macroblock is temporally encoded by an inter-picture motion compensation operation. According to such an operation, a prediction macroblock is identified for the to-be-motion compensated macroblock and is subtracted therefrom to produce a prediction error macroblock. The prediction macroblock originates in another picture, called a reference picture, or may be an interpolation of multiple prediction macroblocks, each originating in different reference pictures. The prediction macroblock need not have precisely the same spatial coordinates (pixel row and column) as the macroblock from which it is subtracted and in fact can be spatially offset therefrom. A motion vector is used to identify the macroblock by its spatial shift and by the reference picture from which it originates. (When the prediction macroblock is an interpolation of multiple prediction macroblocks, a motion vector is obtained for each to-be-interpolated prediction macroblock).
Pictures may be classified as intra or I pictures, predictive or P pictures and bidirectionally predictive or B pictures. An I picture contains only intra macroblocks. A P picture may contain inter macroblocks, but only forward directed predictions from a preceding reference picture are permitted. A P picture can also contain intra macroblocks for which no adequate prediction was found. In addition, a dual prime prediction may be formed for a P picture macroblock in an interlaced picture, which is an interpolated prediction from the immediately two preceding reference fields. A B picture can contain intra macroblocks, inter macroblocks that are forward direction motion compensated, inter macroblocks that are backward directed motion compensated, i.e., predicted from a succeeding reference picture, and inter macroblocks that are bidirectionally motion compensated, i.e., predicted from an interpolation of prediction macroblocks in each of preceding and succeeding reference pictures. If the P or B pictures are interlaced, then each component field macroblock can be separately motion compensated or the two fields can be interleaved to form a frame macroblock and the frame block can be motion compensated at once.
Spatial compression is performed on selected 8×8 luminance pixel blocks and selected 8×8 pixel chrominance blocks of selected prediction error macroblocks, or selected intra macroblocks. Spatial compression includes the steps of discrete cosine transforming each block, quantizing each block, zig-zag (or alternate) scanning each block into a sequence, run-level encoding the sequence and variable length encoding the run-level encoded sequence. Prior to discrete cosine transformation, a macroblock of a frame picture may optionally be formatted as a frame macroblock, including blocks containing alternating lines of samples from each of the two component field pictures of the frame picture, or as a field macroblock, where the samples from different fields are arranged into separate blocks of the macroblock. The quantizer scale factor may be changed on a macroblock-by-macroblock basis and the weighting matrix may be changed on a picture-by-picture basis. Macroblocks, or coded blocks thereof, may be skipped if they have zero (or nearly zero) valued coded data. Appropriate codes are provided into the formatted bitstream of the encoded video signal, such as non-contiguous macroblock address increments, or coded block patterns, to indicate skipped macroblocks and blocks.
Additional formatting is applied to the variable length encoded sequence to aid in identifying the following items within the encoded bitstream: individual sequences of pictures, groups of pictures of the sequence, pictures of a group of pictures, slices (contiguous sequences of macroblocks of a single macroblock row) of pictures, macroblocks of slices and motion vectors and blocks of macroblocks. Some of the above layers are optional, such as the group of pictures layer and the slice layer, and may be omitted from the bitstream if desired. (If slice headers are included in the bitstream, one slice header is provided for each macroblock row.) Various parameters and flags are inserted into the formatted bitstream as well indicating each of the above noted choices (as well as others not described above). The following is a brief list of some of such parameters and flags: picture coding type (I,P,B), macroblock type (i.e., forward predicted, backward predicted, bidirectionally predicted, spatially encoded only) macroblock prediction type (field, frame, dual prime, etc.), DCT type (i.e., frame or field macroblock format for discrete cosine transformation), the quantizer scale code, etc. A repeat_first_field flag may be inserted into the encoded video signal to indicate that a field repeated during a telecine process of converting film frames to NTSC video (using the well known 3:2 pull-down technique) was omitted from the encoded video signal. In addition, error concealment motion vectors optionally may be provided with intra macroblocks for motion compensated recovery of another macroblock in the event the other macroblock is corrupted due to an error.
In encoding the video signal according to MPEG-2, the encoder must produce a bitstream which does not overflow or underflow the buffer of a decoder which decodes the video signal. To that end, the encoder models the decoder's buffer and, in particular, monitors the fullness of the decoder's buffer. The decoder buffer is presumed to fill with bits of the bitstream at a particular rate which is a function of the channel rate at a certain moment of time. Pictures are presumed to be instantly removed at a particular instant relative to the decode and presentation time of each picture. See U.S. patent application Ser. No. 09/084,690 for an in-depth discussion of the modeling of the decoder buffer by an encoder. Using such a model, the encoder can adjust various encoding parameters to control the number of bits produced for each encoded picture in an effort to prevent overflowing or underflowing the decoder's buffer. For example, the encoder can adjust the quantizer scale factor, encourage selection of certain types of encoding over others, add stuffing data to pictures, change the number of B and P pictures, change a threshold quality level used in determining whether to perform intra or inter coding of macroblocks, etc., to increase or reduce the number of bits produced for each picture. Generally speaking, the encoder forms a target bit budget for each picture, which is a function of, among other parameters, the channel rate, the decoder buffer size (normally assumed to be a certain constant), and the vacancy/occupancy of the decoder's buffer immediately before and after removal of the particular picture for which a budget is being generated. The encoder then adjusts its encoding in an attempt to achieve the target bit budget for the picture.
Occasionally, it is desired to re-encode a previously encoded video signal. For example, in some video server or network situations, it is desirable to re-encode the video signal in a fashion other than it was originally encoded to meet network congestion/bandwidth availability constraints, to provide the video signal to different users with varying decoder capability, etc. In another ex
Linzer Elliot
Wells Aaron
Harmonic Inc.
Kelley Chris
Proskauer Rose LLP
Vo Tung
LandOfFree
Video transcoder with bitstream look ahead for rate control... does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Video transcoder with bitstream look ahead for rate control..., we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Video transcoder with bitstream look ahead for rate control... will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2575312