Image analysis – Image compression or coding
Reexamination Certificate
1997-04-30
2001-01-30
Couso, Jose L. (Department: 2721)
Image analysis
Image compression or coding
Reexamination Certificate
active
06181821
ABSTRACT:
BACKGROUND OF THE INVENTION
This invention relates to encoding and multiplexing multiple signals for transmission or storage, and in particular relates to statistically multiplexing of multiple image sequences.
An image sequence encoded at a fixed image quality using a variable-rate encoder produces data at a time varying rate. The data rate, in general, depends on the characteristics of images being encoded, and with some encoding techniques, on the relationships of images being encoded to images nearby in the sequence. For example, images of a fast-moving scene will typically produce a higher data rate than those of a still scene. In addition different encoding techniques, producing different amounts of data, may be used for different images in the sequence. For example, some images may be encoded independently of other images, producing relative more data than other images encoded based on prediction from nearby images. The Motion Picture Expert Group MPEG-2 video encoding standard specifies such a variable-rate image encoding technique.
In addition to producing data at time varying rates at a fixed image quality, some image encoding techniques allow control over an encoding parameter such as a quantization parameter, thereby, indirect control over the image quality and the average output data rate. Reducing the quantization parameter in general reduces average data rate and image quality. For transmission over a capacity constrained channel, variable-rate output from an encoder is stored in a buffer before transmission. The quantization parameter can then be adjusted to control the average data rate. If the buffer contains a large backlog of data, the quantization parameter is reduced thereby resulting in the buffer being filled less quickly by the encoder. If there is a small backlog, the quantization parameter is increased. This type of feedback prevents exceeding the capacity of the buffer or emptying the buffer while using a relatively high image quality.
Multiplexing multiple encoded image sequences on a capacity constrained channel can be done by preallocating portions of the total capacity to individual image sequences. For example, in the case of a fixed-rate channel, each image sequence can be allocated a fixed fraction of the total capacity. However, if the fixed fraction is too small, an encoded image sequence may at times demand a higher data rate than its allocated fraction. If the fixed fraction is too large, there may be times when an encoded image sequence does not need its entire allocated capacity leaving unused channel capacity not available to other image sequences.
When multiple variable-rate encoded image sequences are multiplexed on a single channel, there may be times when one sequence needs a high data rate while another sequence may only need a relatively low data rate. Statistical multiplexing takes advantage of this phenomenon by not allocating fixed portions of the total channel capacity to particular image sequences. The overall sum of the data rates will be time varying. A control mechanism is generally needed to prevent exceeding the capacity of the channel.
FIG. 1
shows a statistical multiplexing arrangement in which image sequences
112
,
114
are encoded by variable-rate image encoders
116
,
118
, fed to a bounded capacity buffer
122
, and multiplexed in multiplexor
120
, before passing through a fixed-data-rate communication path
124
. An encoding controller
130
is responsive to the amount of data in buffer
122
and controls the data rates produced by encoders
116
,
118
in a feedback arrangement in order to keep the amount of data in buffer
122
near a target value and not allowing the amount of data to exceed the capacity of the buffer or allowing the buffer to empty. The encoded data produced by encoders
116
,
118
includes control information sufficient to decode the data even with a time-varying data-rate control.
SUMMARY OF THE INVENTION
In a first aspect, the invention provides an improved technique for variable-rate encoding a signal by processing the signal to determine the data rate demand over a time period for multiple values of an encoding parameter, such as a quantization parameter which affects the encoded data rate, and selecting a value for the encoding parameter with which to encode a segment of the time period. An advantage of the invention over a buffer feedback arrangement is that there may be times when the buffer has a large backlog resulting in feedback control to reduce a quantization parameter, when in fact future image sequences would have naturally produced sufficiently low data rates to allow a larger quantization parameter and therefore, indirectly, a higher image quality to be maintained. Thus, the invention can avoid unnecessary reduction in image quality.
In general, in the first aspect, the invention features encoding a first signal, wherein the encoding produces a first stream of encoded data having a variable data rate generally dependent on values of an encoding parameter, by processing the first signal to determine the data rate demand over a time period for each of a plurality of values of the encoding parameter, selecting a first value of the encoding parameter from the plurality of values based on the determined data rate demands, and encoding the first signal over a segment of the time period using the first value.
Preferred implementations may include one or more of the following features:
The processing, selecting, and encoding steps may be repeated for a plurality of time segments, with different values of the encoding parameter being selected for at least some time segments.
A second signal may be encoded, producing a second stream of encoded data having a variable data rate generally dependent on values of an encoding parameter, and the first and second data streams may be combined to form a combined data stream for transmission or storage, by processing the second signal to determine the data rate demand over the time period for each of a plurality of values of the encoding parameter, selecting a second value of the encoding parameter from the plurality of values based on the determined data rate demands, and encoding the second signal over the segment of the time period using the second value.
The first and second data streams may be multiplexed to form the combined data stream.
A relative delay between the first and the second data streams may be provided to reduce peaks in data rate demand of the combined data stream.
The combined data stream may be stored in a buffer of limited data volume.
The first and second data values may be selected to maintain a desired volume of data in the buffer.
The volume of data currently stored in the buffer may also be used in selecting the values of the encoding parameter.
The first and second signals may be prerecorded image sequences.
The combined data stream may be transmitted over a communication channel of limited data rate capacity.
The step of processing a signal to determine the data rate demand may include presenting each image in the sequence to a variable-rate encoder and recording the amount of data produced for that image for each of a plurality of values of the encoding parameter.
The encoding may include predictive encoding and the time segments over which the same value of the encoding parameter is used may be chosen so that the amounts of data produced by images in one segment are independent of the choice of quantization parameter in another segment.
At least some of the data rate demands may be approximations of the actual data rate demands.
In a second aspect of the invention, the data rate demands of a set of image sequences is determined and a subset of the sequences is selected based on the data rate demands, for example, to satisfy a storage volume constraint. Each of the selected subset of sequences is then encoded for transmission or storage.
In general, the second aspect features encoding for transmission or storage a set of image sequences, by processing the image sequences to determine the data rate demand of
Couso Jose L.
Fish & Richardson P.C.
Massachusetts Institute of Technology
LandOfFree
Predictive source encoding and multiplexing does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Predictive source encoding and multiplexing, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Predictive source encoding and multiplexing will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2551172