Pulse or digital communications – Bandwidth reduction or expansion – Television or motion video signal
Reexamination Certificate
1997-08-08
2002-12-31
Rao, Andy (Department: 2713)
Pulse or digital communications
Bandwidth reduction or expansion
Television or motion video signal
C375S240140, C375S240150
Reexamination Certificate
active
06501795
ABSTRACT:
BACKGROUND OF THE INVENTION
The invention relates to a moving-picture encoding apparatus for encoding a moving picture in block units.
Conventionally, the encoding of a moving picture employs a block encoding method. In block encoding, three encoding methods are used adaptively, namely (1) encoding of interframe difference, (2) motion-compensated encoding of interframe difference and (3) intraframe encoding.
These three methods of encoding are described below.
(1) Encoding of Interframe Difference
This is an encoding method of obtaining the difference between a block to be encoded and a block at the same position in the previous frame and then encoding the difference. With the method of encoding interframe difference, the difference becomes closer to zero the higher the correlation between frames. Encoding is more efficient as a result. The block of difference data is subjected to a discrete cosine transformation, the transform coefficient data obtained is quantized and Huffman codes are subsequently assigned. When all of the data becomes zero after quantization, this means that the quantized image is the same as that of the same block in the previous frame. Accordingly, there is no transmission of code.
(2) Method of Motion-compensated Encoding Interframe Difference
This method involves matching the block to be encoded and neighboring blocks centered on the block at the same position in the previous frame, selecting the block that is most similar and obtaining the difference between the blocks. The difference data is subjected to a discrete cosine transformation and subsequently quantized. Huffman codes are then assigned.
(3) Method of Intraframe Encoding
In a case where the correlation between frames is small, the dynamic range is widened and the amount of information is increased rather than decreased by taking the difference between the frames. In this case, therefore, intraframe encoding is carried out. In intraframe encoding, the discrete cosine transformation is applied to the original image directly, followed by quantization and encoding.
A method of deciding which of these three conventional encoding methods to use is described with reference to
FIG. 6
, which is a flowchart of the deciding processing.
In
FIG. 6
, it is determined at step S
21
whether an output-code buffer (not shown) has overflowed. If the decision rendered is YES, the program proceeds to step S
24
and the method of coding interframe difference is used. If a NO decision is rendered at step S
21
, the program proceeds to step S
22
, where it is decided whether to perform motion compensation. As shown in
FIG. 7
, the decision is made using the average of the absolute-value differences between blocks when motion compensation is performed and the average of the absolute-value differences between blocks when motion compensation is not performed. If the result of the decision at step S
22
is YES (motion compensation ON), the program proceeds to step S
25
and the method of motion-compensated encoding of interframe difference is used.
If the result of the decision at step S
22
is NO (motion compensation OFF), the program proceeds to step S
23
, where a decision is made to use encoding of interframe difference or intraframe encoding. As shown in
FIG. 8
, the decision of step S
22
is performed using variance values of intraframe data and a geometrical average of interframe difference. Intraframe encoding is performed at step S
26
or encoding of intraframe difference at step S
27
, depending upon the decision rendered at step S
23
.
In the above-described decision processing, the statistics of a 16×16 pixel block of luminance data are used. Furthermore, in the encoding described below, the unit of processing is 8×8 pixels. As for the breakdown, the 16×16 pixel block is partitioned into four blocks of 8×8 pixels, and there is a total of six 8×8 pixel blocks of the four 8×8 pixel blocks and two 8×8 pixel blocks of color difference data describing a position that is spatially the same as the four 8×8 pixel blocks. Encoding processing of the same mode is applied to the six 8×8 pixel blocks.
Control of the amount of code generated is described below.
Control for making the amount of generated encoded data conform to the transmission rate is required. This is carried out by the following method: The generated code is fed into a buffer whose code capacity is equivalent to one frame. The code is sequentially transmitted from the buffer at a speed commensurate with the transmission rate. The encoder monitors the sufficiency of the code in the buffer and suppresses the generation of code by enlarging the quantization step when the amount of code that has been generated is excessive. Conversely, the encoder generates more code by reducing the quantization step when the rate at which code is generated falls below the transmission rate. When the buffer overflows, the encoder makes the quantization step very large and halts the generation of code. By virtue of this operation, a fixed amount of code is generated per unit time.
The operation of the encoder is described with reference to FIG.
9
. Block data that has entered from a connector
301
is transmitted on lines
302
and
303
. Meanwhile, an overflow signal from the output-code buffer (not shown) enters from a connector
304
and is transmitted on lines
305
and
306
. A mode decision unit
37
decides the processing mode in accordance with the above-described method upon referring to the block data on line
303
and a previous-frame memory
38
. When it has been decided that the method of encoding interframe difference is appropriate, the mode decision unit
37
reads data of a block at a position identical with that of the input block out of the previous-frame memory
38
via line
307
and outputs the data on the signal line
308
. When it has been decided that the method of motion-compensating encoding of interframe difference is appropriate, the mode decision unit
37
reads the best matching block data out of the previous-frame memory
38
via line
307
and outputs the data on the signal line
308
. When it has been decided that the intraframe encoding method is appropriate, the mode decision unit
37
produces zero data and outputs the data on the line
308
. At the same time, the mode decision unit
37
outputs, on a line
312
, vector data representing the relative position of the best matching block, which has been obtained by motion compensation, and the encoded block. The block data on the line
308
is transmitted on lines
309
and
310
.
The difference between the block data from the line
310
and the input block data from the line
302
is determined and provided to a masking unit
32
via a DCT (direct cosine transform) circuit
31
. When the signal from the connector
304
indicates an overflow state, the mode decision unit
37
selects interframe-difference encoding unconditionally and the masking unit
32
masks all of the difference data to zero. The DCT coefficient data from the masking unit
32
is quantized by a quantizer
33
and the quantized data is provided to an encoder
34
and a reverse quantizer
35
. In response to a selection-mode signal from line
311
, the encoder
34
assigns a Huffman code to the quantized DCT coefficient data and outputs the result on a line
313
. By means of reverse quantization, the reverse quantizer
35
reproduces frequency data identical with that provided to an external decoder, not shown. The reproduced frequency data is again transformed into a difference signal by a reverse-DCT circuit
36
, and this difference signal is added to the signal from the line
309
, thereby reproducing an image identical with that transmitted. This image is again stored in the previous-frame memory
38
.
With the example of the prior art described above, intraframe encoding is performed in a case where the predicted error is very large. Interframe encoding is performed in other cases.
Consider a case in which the 16×16 pixel block of luminance data is
Canon Kabushiki Kaisha
Fitzpatrick ,Cella, Harper & Scinto
Rao Andy
LandOfFree
Adaptable intraframe/interframe motion video encoder does not yet have a rating. At this time, there are no reviews or comments for this patent.
If you have personal experience with Adaptable intraframe/interframe motion video encoder, we encourage you to share that experience with our LandOfFree.com community. Your opinion is very important and Adaptable intraframe/interframe motion video encoder will most certainly appreciate the feedback.
Profile ID: LFUS-PAI-O-2985994