U.S. patent application number 09/880246 was filed with the patent office on 2001-11-29 for reverse coding system for interlaced scanned moving pictures.
This patent application is currently assigned to NEC Corporation. Invention is credited to Oami, Ryoma.
Application Number | 20010046261 09/880246 |
Document ID | / |
Family ID | 16335546 |
Filed Date | 2001-11-29 |
United States Patent
Application |
20010046261 |
Kind Code |
A1 |
Oami, Ryoma |
November 29, 2001 |
Reverse coding system for interlaced scanned moving pictures
Abstract
The invention provides a reversible coding system for an
interlaced scanned moving picture based on reversible discrete
cosine transform which controls selection of frame/field discrete
cosine transform so as to assure a high degree of compatibility
with an ordinary discrete cosine transform coding system. A motion
compensated predictive error is detected from an input signal, and
a frame-scan block and a field-scan block are produced for each
block of the input signal by a frame/field scan production circuit.
One of the blocks is selected by a switch, and reversible discrete
cosine transform is performed for the selected block. Upon the
selection, first and second estimated error calculation circuits
estimate error amounts when reversible discrete cosine transform is
performed for the frame- and field-scan blocks and then inverse
discrete cosine transform is performed for the resultant blocks.
The error amounts are compared with each other, and the lower one
of them is selected by a comparator.
Inventors: |
Oami, Ryoma; (Tokyo,
JP) |
Correspondence
Address: |
OSTROLENK FABER GERB & SOFFEN
1180 AVENUE OF THE AMERICAS
NEW YORK
NY
100368403
|
Assignee: |
NEC Corporation
|
Family ID: |
16335546 |
Appl. No.: |
09/880246 |
Filed: |
June 13, 2001 |
Related U.S. Patent Documents
|
|
|
|
|
|
Application
Number |
Filing Date |
Patent Number |
|
|
09880246 |
Jun 13, 2001 |
|
|
|
09108727 |
Jul 1, 1998 |
|
|
|
6278737 |
|
|
|
|
Current U.S.
Class: |
375/240.2 ;
375/240.24; 375/E7.133; 375/E7.15; 375/E7.163 |
Current CPC
Class: |
H04N 19/112 20141101;
H04N 19/137 20141101; H04N 19/105 20141101 |
Class at
Publication: |
375/240.2 ;
375/240.24 |
International
Class: |
H04N 007/12 |
Foreign Application Data
Date |
Code |
Application Number |
Jul 4, 1997 |
JP |
9-195101 |
Claims
What is claimed is:
1. A reversible moving picture coding system, comprising: means for
determining a motion vector from an interlaced scanned input video
signal; means for producing a motion compensated predictive picture
from a picture for reference using the motion vector; means for
subtracting the motion compensated predictive picture from the
input video signal to produce a predictive error picture; means for
producing a frame-scanned block and a field-scanned block from the
predictive error signal; means for estimating an error amount when
reversible discrete cosine transform is performed for the
frame-scanned block and then ordinary inverse discrete cosine
transform is performed for the resultant block on the decoding side
and outputting the estimated error amount as an estimated error
amount for a frame-scanned block; means for estimating an error
amount when reversible discrete cosine transform is performed for
the field-scanned block and then ordinary inverse discrete cosine
transform is performed for the resultant block on the decoding side
and outputting the estimated error amount as an estimated error
amount for a field-scanned block; means for comparing the estimated
error amount for the frame-scanned block and the estimated error
amount for the field-scanned block with each other to produce a
scan selection signal and outputting the scan selection signal;
means for selecting one of the frame-scanned block and the
field-scanned block in accordance with the scan selection signal
and outputting the selected scanned block as an adaptive-scanned
block; and means for performing reversible discrete cosine
transform for the adaptive-scanned block to produce an
adaptive-scan transform coefficient block.
2. A reversible moving picture coding system, comprising: means for
determining a motion vector from an interlaced scanned input video
signal; means for producing a motion compensated predictive picture
from a picture for reference using the motion vector; means for
subtracting the motion compensated predictive picture from the
input video signal to produce a predictive error picture; means for
producing a first frame-scanned block and a first field-scanned
block from the predictive error picture; means for performing
reversible discrete cosine transform for the first frame-scanned
block to produce a frame-scan transform coefficient block and
outputting the frame-scan transform coefficient block; means for
performing reversible discrete cosine transform for the first
field-scanned block to produce a field-scan transform coefficient
block and outputting the field-scan transform coefficient block;
means for selecting one of the frame-scan transform coefficient
block and the field-scan transform coefficient block in accordance
with a scan selection signal and outputting the selected transform
coefficient block as an adaptive-scan transform coefficient block;
means for performing ordinary inverse discrete cosine transform for
the frame-scan transform coefficient block, locally decoding a
result of the ordinary inverse discrete cosine transform using the
motion vector to produce a frame-scan decoded block and outputting
the frame-scan decoded block; means for performing ordinary inverse
discrete cosine transform for the field-scan transform coefficient
block, locally decoding a result of the ordinary inverse discrete
cosine transform using the motion vector to produce a field-scan
decoded block and outputting the field-scan decoded block; means
for producing a second frame-scanned block and a second
field-scanned block from the input video signal; means for
determining an error between the frame-scan decoded block and the
second frame-scanned block to calculate an error amount for a
frame-scanned block and outputting the error amount for the
frame-scanned block; means for determining an error between the
field-scan decoded block and the second field-scanned block to
calculate an error amount for a field-scanned block and outputting
the error amount for the field-scanned block; and means for
comparing the error amount for the frame-scanned block and the
error amount for the field-scanned block to produce the scan
selection signal and the outputting the scan selection signal.
3. A reversible moving picture coding system, comprising: means for
determining a motion vector from an interlaced scanned input video
signal; means for producing a first motion compensated predictive
picture from a picture for reference using the motion vector; means
for subtracting the first motion compensated predictive picture
from the input video signal to produce a predictive error picture;
means for producing a first frame-scanned block and a first
field-scanned block from the predictive error picture; means for
selecting one of the first frame-scan transform coefficient block
and the first field-scan transform coefficient block in accordance
with a scan selection signal and outputting the selected transform
coefficient block as an adaptive-scan transform coefficient block;
means for performing reversible discrete cosine transform for the
adaptive-scanned block to produce an adaptive-scan transform
coefficient block; means for performing linear operation for the
first frame-scanned block to estimate a predictive error when
reversible transform cosine transform is performed for the first
frame-scanned block and then ordinary inverse discrete cosine
transform is performed for the resultant block on the decoding side
and outputting the predictive error as a frame-scan estimated
predictive error block; means for performing linear operation for
the first field-scanned block to estimate a predictive error when
reversible transform cosine transform is performed for the first
field-scanned block and then ordinary inverse discrete cosine
transform is performed for the resultant block on the decoding side
and outputting the predictive error as a field-scan estimated
predictive error block; means for transforming the adaptive-scan
transform coefficient block by ordinary inverse discrete cosine
transform and producing and outputting a second motion compensated
predictive picture using the motion vector; means for producing a
frame-scan predictive block and a field-scan predictive block from
the second motion compensated predictive picture; means for adding
the frame-scan predictive block to the frame-scan estimated
predictive error block to produce a frame-scan estimated block;
means for adding the field-scan predictive block to the field-scan
estimated predictive error block to produce a field-scan estimated
block; means for producing a second frame-scanned block and a
second field-scanned block from the input video signal; means for
determining an error between the second frame-scanned block and the
frame-scan estimated block and calculating and outputting an
estimated error amount for a frame-scanned block; means for
determining an error between the second field-scanned block and the
field-scan estimated block and calculating and outputting an
estimated error amount for a field-scanned block; and means for
comparing the estimated error amount for the frame-scanned block
and the estimated error amount for the field-scanned block to
produce the scan selection signal and the outputting the scan
selection signal.
4. A reversible moving picture coding system, comprising: means for
determining a motion compensated predictive error from an input
picture signal and producing a frame-scanned block and a
field-scanned block from the motion compensated predictive error
for each block; means for estimating error amounts when reversible
discrete cosine transform is performed for the frame-scanned block
and the field-scanned block and then ordinary inverse discrete
cosine transform is performed for the resultant blocks on the
decoding side and selecting that one of the blocks whose estimated
error amount is the smaller than the other estimated error amount;
and means for performing reversible discrete cosine transform for
the selected scanned block.
Description
BACKGROUND OF THE INVENTION
[0001] 1. Field of the Invention
[0002] The present invention relates to a picture processing
system, and more particularly to reversible transform coding for
interlaced scanned moving pictures.
[0003] 2. Description of the Related Art
[0004] Various reversible coding systems for moving pictures are
conventionally known, and one of reversible coding systems of the
type mentioned has been proposed by the inventor of the invention
of the present application (the reversible coding system is
hereinafter referred to as prior art by the inventor). The prior
art by the inventor is directed to a reversible coding system for
pictures which is based on discrete cosine transform and allows
reversible coding by modifying ordinary discrete cosine transform.
A reversible moving picture coding system can be implemented by
applying the reversible coding system for pictures to a coding
algorithm of the MPEG-2 (Moving Picture Experts Group-Phase 2).
[0005] MPEG-2 coding originally is non-reversible coding. However,
MPEG-2 coding is non-reversible only at the quantization part
thereof. The entire coding can be made reversible coding by
replacing the discrete cosine transform and quantization parts with
reversible discrete cosine transform. In the prior art by the
inventor, a reversible moving picture coding system is implemented
by the countermeasure just described.
[0006] FIG. 7 shows in block diagram a construction of an encoder
which effects reversible moving picture coding.
[0007] Referring to FIG. 7, a motion estimation circuit 5 effects
motion estimation between a picture coded in the past and a picture
of an object of coding and outputs a motion vector for each macro
block. A predictor 4 produces a motion compensated predictive
picture from the reference picture in accordance with motion
vectors outputted from the motion estimation circuit 5 and outputs
the motion compensated predictive picture. An adder 3 subtracts the
motion compensated predictive picture outputted from the predictor
4 from the coding object picture to produce a predictive error
picture. A reversible discrete cosine transform (lossless DCT;
LDCT) circuit 1 performs reversible discrete cosine transform for
the predictive error picture outputted from the adder 3 for each
block and outputs transform coefficients.
[0008] It is to be noted that the encoder shown in FIG. 7 does not
include a local decoding loop which is included in an ordinary
MPEG-2 encoder. This is because, in reversible coding, a locally
decoded picture coincides completely with the original picture and
no local decoding is required.
[0009] In the MPEG-2, an object of coding is an interlaced scanned
moving picture. When discrete cosine transform is performed, either
field scanned blocks, that is, blocks obtained by interlaced
scanning, or frame-scanned blocks, that is, progressively scanned
blocks, can be selected. As a selection controlling method for such
selection, for example, a method of the MPEG-2 TM-5 (Test Model 5)
or another method disclosed in Japanese Patent Laid-Open
Application No. Heisei 8-102950 (title of the invention: Coding
Apparatus and Motion Discrimination Method) is known.
[0010] In the MPEG-2 TM-5, a correlation coefficient of pixel
values is determined for each macro block between fields to effect
selection of frame-scanning or field-scanning. The magnitude of the
correlation coefficient is discriminated with a threshold value,
and if the correlation coefficient is higher than the threshold
value, then frame-scanning is selected, but in any other case,
field-scanning is selected to effect discrete cosine transform.
[0011] In the coding apparatus disclosed in Japanese Patent
Laid-Open Application No. Heisei 8-102950 mentioned above, a
difference in pixel value is determined for each block between
fields, and selection between frame discrete cosine transform and
field discrete cosine transform is controlled based on an absolute
value of a sum when such differences are added in a vertical
direction. More particularly, an absolute value of a difference sum
described above is determined for each column first. Then, the
absolute values of the difference sums of the columns are added to
determine a sum total, and the sum total is compared with the
threshold value. Simultaneously, a maximum value among the absolute
values of the difference sums of the columns is determined, and the
maximum value is compared with the threshold value. Then, if both
of the sum total and the maximum value are higher than the
threshold value, then field discrete cosine transform is selected,
but in any other case, frame discrete cosine transform is
selected.
[0012] The selection control described above can be incorporated
into the system shown in FIG. 7. In particular, frame discrete
cosine transform and field discrete cosine transform can be
adaptively changed over using the prior art by the inventor. The
applied system is shown in FIG. 8.
[0013] Referring to FIG. 8, a frame/field adaptive-scanned block
production circuit 6 adaptively selects field-scanning or frame
scanning based on a conventional system from a predictive error
picture outputted from the adder 3 to produce a block and outputs
the thus produced block.
[0014] However, even if the system of the MPEG-2 TM-5 or the system
disclosed in Japanese Patent Laid-Open Application No. Heisei
8-120950 mentioned hereinabove is combined with the prior art by
the inventor, a desired result cannot necessarily be obtained from
the point of view of compatibility between a reversible discrete
cosine transform coding system and a conventional discrete cosine
transform coding system. Here, the compatibility signifies a matter
of which degree of picture quality is obtained when a picture coded
using reversible discrete cosine transform is decoded by ordinary
inverse discrete cosine transform. It is considered that the higher
the picture quality obtained then, the superior in compatibility
the reversible discrete cosine transform coding system.
[0015] The reason why a desirable result cannot be obtained is that
the conventional adaptive selection system between frame discrete
cosine transform and field discrete cosine transform is optimized
in terms of reduction of the code amount where ordinary discrete
cosine transform is used, and does not take augmentation in
compatibility into consideration. Conversely speaking, the
selection system can be re-examined from the point of view of
augmentation of the compatibility.
SUMMARY OF THE INVENTION
[0016] It is an object of the present invention to provide a
reversible moving picture- coding system wherein, where reversible
discrete cosine transform is applied to the MPEG-2, selection
control of frame/field discrete cosine transform is modified to
augment the compatibility between a reversible discrete cosine
transform coding system and a conventional discrete cosine
transform coding system. More particularly, the object of the
present invention is to provide a reversible moving picture coding
system by which a decoded signal of an augmented quality is
obtained when a picture coded using reversible discrete cosine
transform is decoded using ordinary inverse discrete cosine
transform or when a picture coded using ordinary discrete cosine
transform is decoded using reversible inverse discrete cosine
transform.
[0017] In order to attain the object of the present invention,
according to an aspect of the present invention, there is provided
a reversible moving picture coding system, comprising means for
determining a motion vector from an interlaced scanned input video
signal, means for producing a motion compensated predictive picture
from a picture for reference using the motion vector, means for
subtracting the motion compensated predictive picture from the
input video signal to produce a predictive error picture, means for
producing a frame-scanned block and a field-scanned block from the
predictive error signal, means for estimating an error amount when
reversible discrete cosine transform is performed for the
frame-scanned block and then ordinary inverse discrete cosine
transform is performed for the resultant block on the decoding side
and outputting the estimated error amount as an estimated error
amount for a frame-scanned block, means for estimating an error
amount when reversible discrete cosine transform is performed for
the field-scanned block and then ordinary inverse discrete cosine
transform is performed for the resultant block on the decoding side
and outputting the estimated error amount as an estimated error
amount for a field-scanned block, means for comparing the estimated
error amount for the frame-scanned block and the estimated error
amount for the field-scanned block with each other to produce a
scan selection signal and outputting the scan selection signal,
means for selecting one of the frame-scanned block and the
field-scanned block in accordance with the scan selection signal
and outputting the selected scanned block as an adaptive-scanned
block, and means for performing reversible discrete cosine
transform for the adaptive-scanned block to produce an
adaptive-scan transform coefficient block.
[0018] According to another aspect of the present invention, there
is provided a reversible moving picture coding system, comprising
means for determining a motion vector from an interlaced scanned
input video signal, means for producing a motion compensated
predictive picture from a picture for reference using the motion
vector, means for subtracting the motion compensated predictive
picture from the input video signal to produce a predictive error
picture, means for producing a first frame-scanned block and a
first field-scanned block from the predictive error picture, means
for performing reversible discrete cosine transform for the first
frame-scanned block to produce a frame-scan transform coefficient
block and outputting the frame-scan transform coefficient block,
means for performing reversible discrete cosine transform for the
first field-scanned block to produce a field-scan transform
coefficient block and outputting the field-scan transform
coefficient block, means for selecting one of the frame-scan
transform coefficient block and the field-scan transform
coefficient block in accordance with a scan selection signal and
outputting the selected transform coefficient block as an
adaptive-scan transform coefficient block, means for performing
ordinary inverse discrete cosine transform for the frame-scan
transform coefficient block, locally decoding a result of the
ordinary inverse discrete cosine transform using the motion vector
to produce a frame-scan decoded block and outputting the frame-scan
decoded block, means for performing ordinary inverse discrete
cosine transform for the field-scan transform coefficient block,
locally decoding a result of the ordinary inverse discrete cosine
transform using the motion vector to produce a field-scan decoded
block and outputting the field-scan decoded block, means for
producing a second frame-scanned block and a second field-scanned
block from the input video signal, means for determining an error
between the frame-scan decoded block and the second frame-scanned
block to calculate an error amount for a frame-scanned block and
outputting the error amount for the frame-scanned block, means for
determining an error between the field-scan decoded block and the
second field-scanned block to calculate an error amount for a
field-scanned block and outputting the error amount for the
field-scanned block, and means for comparing the error amount for
the frame-scanned block and the error amount for the field-scanned
block to produce the scan selection signal and the outputting the
scan selection signal.
[0019] According to a further aspect of the present invention,
there is provided a reversible moving picture coding system,
comprising means for determining a motion vector from an interlaced
scanned input video signal, means for producing a first motion
compensated predictive picture from a picture for reference using
the motion vector, means for subtracting the first motion
compensated predictive picture from the input video signal to
produce a predictive error picture, means for producing first
frame-scanned blocks and first field-scanned blocks from the
predictive error picture, means for selecting one of the first
frame-scan transform coefficient block and the first field-scan
transform coefficient block in accordance with a scan selection
signal and outputting the selected transform coefficient block as
an adaptive-scan transform coefficient block, means for performing
reversible discrete cosine transform for the adaptive-scanned block
to produce an adaptive-scan transform coefficient block, means for
performing linear operation for the first frame-scanned block to
estimate a predictive error when the first frame-scanned block is
reversible transform cosine transformed and then ordinary inverse
discrete cosine transformed on the decoding side and outputting the
predictive error as a frame-scan estimated predictive error block,
means for performing linear operation for the first field-scanned
block to estimate a predictive error when reversible transform
cosine transform is performed for the first field-scanned block and
then ordinary inverse discrete cosine transform is performed for
the resultant block on the decoding side and outputting the
predictive error as a field-scan estimated predictive error block,
means for transforming the adaptive-scan transform coefficient
block by ordinary inverse discrete cosine transform and producing
and outputting a second motion compensated predictive picture using
the motion vector, means for producing a frame-scan predictive
block and a field-scan predictive block from the second motion
compensated predictive picture, means for adding the frame-scan
predictive block to the frame-scan estimated predictive error block
to produce a frame-scan estimated block, means for adding the
field-scan predictive block to the field-scan estimated predictive
error block to produce a field-scan estimated block, means for
producing a second frame-scanned block and a second field-scanned
block from the input video signal, means for determining an error
between the second frame-scanned block and the frame-scan estimated
block and calculating and outputting an estimated error amount for
a frame-scanned block, means for determining an error between the
second field-scanned block and the field-scan estimated block and
calculating and outputting an estimated error amount for a
field-scanned block, and means for comparing the estimated error
amount for the frame-scanned block and the estimated error amount
for the field-scanned block to produce the scan selection signal
and the outputting the scan selection signal.
[0020] According to a still further aspect of the present
invention, there is provided a reversible moving picture coding
system, comprising means for determining a motion compensated
predictive error from an input picture signal and producing a
frame-scanned block and a field-scanned block from the motion
compensated predictive error for each block, means for estimating
error amounts when reversible discrete cosine transform is
performed for the frame-scanned block and the field-scanned block
and then ordinary inverse discrete cosine transform is performed
for the resultant blocks on the decoding side and selecting that
one of the blocks whose estimated error amount is the smaller than
the other estimated error amount, and means for performing
reversible discrete cosine transform for the selected scanned
block.
[0021] With the reversible moving picture coding systems, since
selection between frame discrete cosine transform and field
discrete cosine transform is modified to a method which is
preferable for the compatibility between a reversible discrete
cosine transform coding system and a conventional discrete cosine
transform coding system, an advantage that the compatibility is
augmented can be achieved. More particularly, with the reversible
moving picture coding system, even when a picture coded using
reversible discrete cosine transform is decoded using conventional
inverse discrete cosine transform, a higher picture quality than
those achieved by conventional systems can be achieved.
[0022] The above and other objects, features and advantages of the
present invention will become apparent from the following
description and the appended claims, taken in conjunction with the
accompanying drawings in which like parts or elements are denoted
by like reference symbols.
BRIEF DESCRIPTION OF THE DRAWINGS
[0023] FIG. 1 is a block diagram showing a construction of a
reversible moving picture coding system to which the present
invention is applied;
[0024] FIG. 2 is a similar view but showing a construction of
another reversible moving picture coding system to which the
present invention is applied;
[0025] FIG. 3 is a similar view but showing a construction of a
further reversible moving picture coding system to which the
present invention is applied;
[0026] FIG. 4 is a block diagram showing an example of a
construction of first and second local decoding circuits of the
reversible moving picture coding system of FIG. 2;
[0027] FIG. 5 is a block diagram showing an example of a
construction of a predictive picture production unit of a further
reversible moving picture coding system to which the present
invention is applied;
[0028] FIG. 6 is a graph illustrating augmentation of the picture
quality of a decoded picture when reversible discrete cosine
transform is used for coding and ordinary inverse discrete cosine
transform is used for decoding in accordance with the present
invention;
[0029] FIG. 7 is a block diagram showing a construction of an
encoder which employs reversible discrete cosine transform; and
[0030] FIG. 8 is a block diagram showing an encoder in which a
conventional selection circuit which adaptively selects one of
frame discrete cosine transform and field discrete cosine transform
is incorporated.
DESCRIPTION OF THE PREFERRED EMBODIMENTS
[0031] Referring first to FIG. 1, there is shown in block diagram a
circuit construction of a reversible moving picture coding system
to which the present invention is applied. A motion estimation
circuit 5 effects motion estimation for an input video signal and
outputs motion vectors. A predictor 4 produces a motion compensated
predictive picture from a reference picture using the motion
vectors determined by the motion estimation circuit 5. An adder 3
subtracts the motion compensated predictive picture from a coding
object picture of the input video signal to produce a predictive
error picture. A frame/field scan production circuit. 10 produces
frame-scanned blocks and field-scanned blocks from the predictive
error picture.
[0032] A first estimated error amount calculation circuit 11
estimates, from the frame-scanned block, an error when the block is
coded using reversible discrete cosine transform and then decoded
using ordinary inverse discrete cosine transform, and calculates an
estimated error amount of the frame-scanned block from the error. A
second estimated error amount calculation circuit 12 estimates,
from the field-scanned block, an error when the block is coded
using reversible discrete cosine transform and then decoded using
ordinary inverse discrete cosine transform, and calculates an
estimated error amount of the field-scanned block from the error. A
comparator 13 compares the estimated error amount of the
frame-scanned block and the estimated error amount of the
field-scanned block with each other and outputs a scan selection
signal. A switch 14 selects one of the frame-scanned block and the
field-scanned block using the scan selection signal as a switching
control signal. Then, a reversible discrete cosine transform
(lossless DCT; LDCT) circuit 1 performs reversible discrete cosine
transform for the selected block and outputs transform
coefficients.
[0033] In operation, an input video signal is divided in frames,
and such frames are inputted in an order in which they are to be
coded. A coding object picture of the input video signal is
inputted to the motion estimation circuit 5. The motion estimation
circuit 5 performs motion compensation for the coding object
picture with a reference picture to calculate motion vectors and
outputs the motion vectors.
[0034] Using the motion vectors, the predictor 4 produces a motion
compensated predictive picture from the reference picture. Then,
the predictive picture is subtracted from the coding object picture
by the adder 3 to produce a predictive error picture. The
predictive error picture is inputted to the frame/field scan
production circuit 10, by which frame-scanned blocks and
field-scanned blocks are produced for each macro block.
[0035] The following processes are performed in units of a block. A
frame-scanned block and a field-scanned block are inputted to the
first estimated error amount calculation circuit 11 and the second
estimated error amount calculation circuit 12, by which an
estimated error amount for the frame-scanned block and an estimated
error amount for the field-scanned block are calculated,
respectively. The estimated error amounts are estimated amounts of
errors when the blocks are coded using reversible discrete cosine
transform and then decoded using ordinary inverse discrete cosine
transform. In the following, a calculation method for them is
described.
[0036] Where a matrix obtained by normalizing row vectors of a
transform matrix of reversible discrete cosine transform is
represented by A and a matrix composed of pixel values of a block
is represented by X, a transform coefficient obtained by reversible
discrete cosine transform of the block has a value proximate to the
following expression (1):
AXA.sup.T (1)
[0037] Here, where the original matrix representative of discrete
cosine transform is represented by C, the following expression
(2)
C.sup.-1AXA.sup.T(C.sup.-1).sup.T=C.sup.-1AX(C.sup.-1A).sup.T
(2)
[0038] represents a result of ordinary inverse discrete cosine
transform performed for the expression (1) above.
[0039] Consequently, the following expression (3)
C.sup.-1AX(C.sup.-1A).sup.T-X (3)
[0040] is considered to be approximate to an error when reversible
discrete cosine transform is performed and then ordinary inverse
discrete cosine transform is performed for the block X.
[0041] Thus, an estimated error amount can be calculated from the
value of the expression (3) above. If, for example, a mean-square
error of the block is used as the error amount, then a mean-square
of the individual elements of the expression (3) above should be
determined to calculate an estimated error amount.
[0042] The estimated error amount for the field-scanned block and
the estimated error amount the frame-scanned block calculated in
this manner are inputted to the comparator 13, by which the values
thereof are compared with each other. The comparator 13 thus
outputs a signal representative of a relationship in magnitude
between the estimated error amounts, that is, a scan selection
signal. In accordance with the scan selection signal, one of the
frame-scanned block and the field-scanned block is selected by the
switch 14. Here, that one of the blocks whose estimated error
amount is determined to be smaller as a result of the comparison of
the comparator 13 is selected.
[0043] For the block selected by the switch 14, reversible discrete
cosine transform is performed by the reversible discrete cosine
transform circuit 1 to determine a transform coefficient. Transform
coefficients of the individual blocks are calculated in this
manner. It is to be noted that information representative of which
one of frame-scanning and field-scanning is selected for each block
is coded as header information.
[0044] As described above, in the reversible moving picture coding
system of the first embodiment of the present invention, an error
when reversible discrete cosine transform is used for coding and
ordinary inverse discrete cosine transform is used for decoding is
estimated, and that scanning which is estimated to provide the
smaller error is selected for each block to effect coding of the
block. Consequently, the compatibility between reversible discrete
cosine transform coding and ordinary discrete cosine transform
coding is augmented.
[0045] FIG. 2 shows in block diagram a circuit construction of
another reversible moving picture coding system to which the
present invention is applied. Referring to FIG. 2, the reversible
moving picture coding system shown includes a motion estimation
circuit 5, a predictor 4 and an adder 3 which are all similar to
those of the reversible moving picture coding system of the first
embodiment described above with reference to FIG. 1. Thus,
overlapping description of the common components is omitted here to
avoid redundancy.
[0046] A first frame/field scan production circuit 30 produces
frame-scanned blocks and field-scanned blocks from a predictive
error picture outputted from the adder 3. A first reversible
discrete cosine transform circuit 32 performs reversible discrete
cosine transform for the frame-scanned block and outputs a
frame-scan transform coefficient block. A second reversible
discrete cosine transform circuit 33 performs reversible discrete
cosine transform for the field-scanned block and outputs a
field-scan transform coefficient block. A first local decoder 34
decodes the frame-scan transform coefficient block using a motion
vector determined by the motion estimation circuit 5 and outputs a
frame-scan decoded block. A second local decoder 35 decodes the
field-scan transform coefficient block using the motion vector
determined by the motion estimation circuit 5 and outputs a
field-scan decoded block.
[0047] A second frame/field scan production circuit 31 produces
frame-scanned blocks and field-scanned blocks from a coding object
picture of an input video signal. A first error amount calculation
circuit 36 determines an error between the frame-scanned block
outputted from the second frame/field scan production circuit 31
and the frame-scan decoded block to calculate an error amount and
outputs the error amount. A second error amount calculation circuit
37 determines an error between the field-scanned block outputted
from the second frame/field scan production circuit 31 and the
field-scan decoded block to calculate an error amount and outputs
the error amount.
[0048] A comparator 38 compares the error amounts outputted from
the first error amount calculation circuit 36 and the second error
amount calculation circuit 37 with each other and outputs a scan
selection signal. A switch 39 selects one of the frame-scan
transform coefficient block and the field-scan transform
coefficient block.
[0049] In operation, motion vectors are calculated from an input
video signal by the motion estimation circuit 5, and a motion
compensated predictive picture is produced by the predictor 4.
Then, the predictive picture is subtracted from a coding object
picture by the adder 3 to produce a predictive error picture. The
operations just described are same as those of the reversible
moving picture coding system of the first embodiment described
hereinabove with reference to FIG. 1.
[0050] From the predictive error picture produced in such a manner
as described above, frame-scanned blocks and field-scanned blocks
are produced by the first frame/field scan production circuit 30 in
a similar manner as by the frame/field scan production circuit 10
shown in FIG. 1. The following processes are performed in units of
a block.
[0051] For the frame-scanned block and the field-scanned block,
reversible discrete cosine transform is performed by the first
reversible discrete cosine transform circuit 32 and the second
reversible discrete cosine transform circuit 33 so that a
frame-scan transform coefficient block and a field-scan transform
coefficient block are produced by them, respectively. Then, one of
the two transform coefficient blocks is selected in accordance with
a scan selection signal which will be hereinafter described by the
switch 39 and is outputted as an adaptive-scan transform
coefficient block.
[0052] Now, production of the scan selection signal is described.
First, the frame-scan transform coefficient block and the
field-scan transform coefficient block outputted from the first
reversible discrete cosine transform circuit 32 and second
reversible discrete cosine transform circuit 33 are inputted to the
first local decoder 34 and the second local decoder 35, by which a
frame-scan decoded block and a field-scan decoded block are
produced, respectively. The first and second local decoders 34 and
35 have such a common construction as shown in FIG. 4.
[0053] Referring to FIG. 4, a transform coefficient block inputted
to the local decoder shown is first dequantized by a dequantizer
100 and then inverse discrete cosine transform is performed for the
resultant block by an inverse discrete cosine transform (IDCT)
circuit 101. Then, a predictive block outputted from a predictor
103 is added to a result of the inverse discrete cosine transform
of the inverse discrete cosine transform circuit 101 by an adder
102 to produce a decoded block, and the thus produced decoded block
is outputted from the local decoder. The predictor 103 produces the
predictive block from a locally decoded reference picture stored
therein using the motion vector outputted from the motion
estimation circuit 5.
[0054] In this instance, if the input to the local decoder is a
frame-scanned block, then the local decoder produces a frame-scan
predictive block, but if the input to the local decoder is a
field-scan block, then the local decoder produces a field-scan
predictive block.
[0055] It is to be noted that the locally decoded reference
pictures stored in the first local decoder 34 and the second local
decoder 35 shown in FIG. 2 are the same reference picture. In order
to produce the local decoded reference picture, one of a frame-scan
decoded block and a field-scan decoded block is selected for each
block. Here, which one of them should be selected depends upon the
scan selection signal outputted from the comparator 38, and the
same one of them that is selected by the switch 39 is selected.
Then, the thus produced locally decoded reference picture is used
to encode a video signal that will be inputted later.
[0056] Referring to FIG. 2, the frame-scan decoded block and the
field-scan decoded block outputted from the first local decoder 34
and the second local decoder 35 are inputted to the first error
amount calculation circuit 36 and the second error amount
calculation circuit 37, respectively. Meanwhile, the second
frame/field scan production circuit 31 produces a frame-scanned
block and a field-scanned block from the coding object picture of
the input video signal in a similar manner as in the first
frame/field scan production circuit 30. The frame-scanned block and
the field-scanned block are inputted to the first error amount
calculation circuit 36 and the second error amount calculation
circuit 37, respectively. Each of the first and second error amount
calculation circuits 36 and 37 calculates a difference between the
two blocks inputted thereto and further calculates an error amount.
For example, where a mean-square error is used as the error amount,
each of the first and second error amount calculation circuits 36
and 37 calculates and outputs a square-mean between the blocks as
an error amount.
[0057] The error amounts outputted from the first and second error
amount calculation circuits 36 and 37 are inputted to the
comparator 38, by which they are compared with each other to
discriminate which is larger. The comparator 38 thus outputs a
signal representative of the smaller one of the error amounts as a
scan selection signal. Then, the block which exhibits the smaller
error is selected in accordance with the scan selection signal as
described above by the switch 39.
[0058] As described above, in the reversible moving picture coding
apparatus of the second embodiment of the present invention, for
each block, reversible discrete cosine transform is actually
performed for both of a frame-scanned block and a field-scanned
block and that scanning which exhibits the higher quality of a
decoded picture is selected. Accordingly, optimum selection can be
performed from the point of view of the compatibility.
[0059] FIG. 3 shows in block diagram a circuit construction of a
further reversible moving picture coding system to which the
present invention is applied. Referring to FIG. 3, the reversible
moving picture coding system shown includes a motion estimation
circuit 5, a predictor 4, an adder 3, a reversible discrete cosine
transform circuit 1 and a switch 14 which are all similar to those
of the reversible moving picture coding system described
hereinabove with reference to FIG. 1. The reversible moving picture
roding system further includes a first frame/field scan production
circuit 30, a second frame/field scan production circuit 31 and a
comparator 38 which are all similar to those of the reversible
moving picture coding system described hereinabove with reference
to FIG. 2. Thus, overlapping description of such common components
as those mentioned above is omitted here to avoid redundancy.
[0060] A predictive picture production section 60 produces a
predictive picture from a transform coefficient block outputted
from the reversible discrete cosine transform circuit 1 using
motion vectors outputted from the motion estimation circuit 5.
[0061] A third frame/field scan production circuit 59 produces and
outputs a frame-scan predictive block and a field-scan predictive
block from the predictive picture outputted from the predictive
picture production section 60.
[0062] A first decoded-predictive-error estimation circuit 52
estimates a predictive error when reversible discrete cosine
transform is performed for the frame-scanned block outputted from
the first frame/field scan production circuit 30 and then ordinary
inverse discrete cosine transform is performed for the resultant
block on the decoding side, and outputs a frame-scan estimated
predictive error block.
[0063] A second decoded-predictive-error estimation circuit 53
estimates a predictive error when reversible discrete cosine
transform is performed for the field-scanned block outputted from
the first frame/field scan production circuit 30 and then ordinary
inverse discrete cosine transform is performed for the resultant
block on the decoding side, and outputs a field-scan estimated
predictive error block.
[0064] An adder 54 adds the frame-scan predictive block and the
frame-scan estimated predictive error block.
[0065] Another adder 55 adds the field-scan predictive block and
the field-scan estimated predictive error block.
[0066] A first error amount calculation circuit 56 calculates an
error amount from the field-scanned block outputted from the second
frame/field scan production circuit 31 and an output of the adder
54.
[0067] A second error amount calculation circuit 57 calculates an
error amount from the frame-scanned block outputted from the second
frame/field scan production circuit 31 and an output of the adder
55.
[0068] In operation, motion vectors are calculated from an input
picture signal by the motion estimation circuit 5, and a motion
compensated predictive picture is produced by the predictor 4 and
is subtracted from a coding object picture by the adder 3 to
produce a predictive error picture. The operations just described
are similar to those of the reversible moving picture coding system
described hereinabove with reference to FIG. 1.
[0069] From the thus produced predictive error picture, a
frame-scanned block and a field-scanned block are produced by the
first frame/field scan production circuit 30 in a similar manner as
in the frame/field scan production circuit 10 described above with
reference to FIG. 1. The later processing is performed in units of
a block.
[0070] One of the frame-scanned block and the field-scanned block
outputted from the first frame/field scan production circuit 30 is
selected in accordance with a scan selection signal by the switch
14. Further, reversible discrete cosine transform is performed for
the selected scanned block by the reversible discrete cosine
transform circuit 1. Consequently, an adaptive-scan transform
coefficient block is produced by and outputted from the reversible
discrete cosine transform circuit 1. The operations just described
are similar to those in the reversible moving picture coding system
described hereinabove with reference to FIG. 1.
[0071] Now, production of the scan selection signal used by the
switch 14 is described. First, the frame-scanned block and the
field-scanned block outputted from the first frame/field scan
production circuit 30 are inputted to the first
decoded-predictive-error estimation circuit 52 and the second
decoded-predictive-error estimation circuit 53, respectively. Then,
a decoded-predictive-error is estimated by each of the first
decoded-predictive-error estimation circuit 52 and the second
decoded-predictive-error estimation circuit 53, and a frame-scan
estimated predictive error block and a field-scan estimated
predictive error block are outputted from them, respectively. Here,
each of the first and second decoded-predictive-error estimation
circuits 52 and 53 estimates a predictive error when reversible
discrete cosine transform is performed on the coding side and
ordinary inverse discrete cosine transform is performed on the
decoding side.
[0072] As described above, a value obtained by performing
reversible discrete cosine transform for a block X and then
performing ordinary inverse discrete cosine transform for the block
X is a value proximate to the expression (2) given hereinabove.
Consequently, the present circuit performs calculation represented
by the expression (2) above for an inputted block to calculate an
estimated predictive error block.
[0073] Meanwhile, the adaptive-scan transform coefficient block
outputted from the reversible discrete cosine transform circuit 1
is inputted to the predictive picture production section 60. The
predictive picture production section 60 produces, based on the
motion vectors outputted from the motion estimation circuit 5, a
motion compensated predictive picture which is considered to be
obtained on the decoding side.
[0074] FIG. 5 shows in block diagram a construction of the
predictive picture production section 60. Referring to FIG. 5, in
the predictive picture production section 60 shown, an
adaptive-scan transform coefficient block inputted is first
dequantized by a dequantizer (IQ) 130 and then inverse discrete
cosine transform is performed for the resultant block by an inverse
discrete cosine transform circuit (IDCT) 131. To a result obtained
by the inverse discrete cosine transform, an output of a predictor
133 is added by an adder 132 to obtain a locally decoded picture.
The locally decoded picture obtained is stored into the predictor
133. The predictor 133 performs motion compensation using a locally
decoded picture stored already therein as a reference picture to
produce a motion compensated predictive picture. Then, the motion
compensated predictive picture is added to the result of the
inverse discrete cosine transform by the adder 132 and
simultaneously inputted to the third frame/field scan production
circuit 59 shown in FIG. 3.
[0075] From the predictive picture produced by performing motion
compensation for the local decoded picture stored already in this
manner, a frame-scan predictive block and a field-scan predictive
block are produced by the third frame/field scan production circuit
59.
[0076] Referring back to FIG. 3, the frame-scan predictive block
and the field-scan predictive block are added to the frame-scan
estimated predictive error block and the field-scan estimated
predictive error block by the adders 54 and 55 to obtain a
frame-scan estimated block and a field-scan estimated block,
respectively. The frame-scan estimated block and the field-scan
estimated block are inputted to the first error amount calculation
circuit 56 and the second error amount calculation circuit 57,
respectively.
[0077] Meanwhile, a frame-scanned block and a field-scanned block
are produced from the input video signal by the second frame/field
scan production circuit 31 and are inputted to the first error
amount calculation circuit 56 and the second error amount
calculation circuit 57, respectively.
[0078] Operations of the first error amount calculation circuit 56
and the second error amount calculation circuit 57 are similar to
those of the first error amount calculation circuit 36 and the
second error amount calculation circuit 37 described hereinabove
with reference to FIG. 2. Outputs of the first error amount
calculation circuit 56 and the second error amount calculation
circuit 57 are compared in magnitude by the comparator 38, and a
scan selection signal is outputted from the comparator 38 in a
similar manner as in the reversible moving picture coding system
described hereinabove with reference to FIG. 2.
[0079] Then, one of the frame-scanned block and the field-scanned
block is selected in accordance with the scan selection signal by
the switch 14 and is outputted as an adaptive-scanned block as
described above.
[0080] The reversible moving picture coding system of the third
embodiment described above can achieve reduction of the amount of
calculation comparing with that of the reversible moving picture
coding system of the second embodiment described hereinabove.
[0081] In the reversible moving picture coding system of the second
embodiment described hereinabove, reversible discrete cosine
transform must be performed for both of a frame-scanned block and a
field-scanned block. Since reversible discrete cosine transform
requires a larger amount of calculation than ordinary discrete
cosine transform, it is preferable to minimize the number of times
by which reversible discrete cosine transform is performed.
[0082] The reversible moving picture coding system of the third
embodiment reduces the amount of calculation by approximating
reversible discrete cosine transform with linear operation in place
of actually performing reversible discrete cosine transform.
[0083] The reversible moving picture coding system of the third
embodiment exhibits a result somewhat inferior to that of the
reversible moving picture coding system of the second embodiment
described hereinabove from the point of view of the compatibility
since selection control therein partly involves approximation as
described above. However, since selection control is performed
using a local decoded picture, a result preferable in that the
degree of approximation is lower than that of the reversible moving
picture coding system of the first embodiment described hereinabove
is obtained.
[0084] The reversible moving picture coding systems of the first to
third embodiments of the present invention described above are
described in more detail below. In the reversible moving picture
coding systems of the first and third embodiments, linear
calculation regarding a matrix (C.sup.-1A) is performed as given by
the expressions (3) and (2). Where the transform matrix of
reversible discrete cosine transform is the following expression
(4), 1 [ 1 1 1 1 1 1 1 1 5 4 3 1 - 1 - 3 - 4 - 5 12 5 - 5 - 12 - 12
- 5 5 12 4 - 1 - 5 - 3 3 5 1 - 4 1 - 1 - 1 1 1 - 1 - 1 1 3 - 5 1 4
- 4 - 1 5 - 3 5 - 12 12 - 5 - 5 12 - 12 5 1 - 3 4 - 5 5 - 4 3 - 1 ]
( 4 )
[0085] the matrix (C.sup.-1A) is given by the following expression
(5): 2 [ .9996 - .0129 .0055 - .0132 .0132 - .0065 .0140 .0004 -
.0065 .9996 .0132 .0140 - .0129 - .0132 .0004 .0055 .0140 - .0312
.9996 .0055 - .0065 .0004 .0132 - .0129 .0132 .0055 .0140 .9996
.0004 - .0129 - .0065 - .0132 - .0132 - .0065 - .0129 .0004 .9996
.0140 .0055 .0132 - .0129 .0132 .0004 - .0065 .0055 .9996 - .0132
.0140 .0055 .0004 - .0132 - .0129 .0140 .0132 .9996 - .0065 .0004
.0140 - .0065 .0132 - .0132 .0055 - .0129 .9996 ] ( 5 )
[0086] Consequently, where the transform matrix of reversible
discrete cosine transform is the expression (4) above, the matrix
given by the expression (5) above is used for estimation of an
error.
[0087] As an example, a result of an examination regarding the
compatibility performed using the reversible moving picture coding
systems of the embodiments of the present invention described above
is described. An error of a decoded picture where reversible
discrete cosine transform was used for coding and conventional
discrete cosine transform was used for decoding was examined.
[0088] A picture signal used was of the 720.times.480 pixels, the
4:2:2 format and the 150 frames. For coding, basically an algorithm
of the MPEG-2 was used with the discrete cosine transform thereof
replaced by the reversible discrete cosine transform described
above. Further, transform coefficients obtained by the reversible
discrete cosine transform were not quantized, or in other words,
the quantization step size was 1. In this instance, the
specifications of the MPEG-2 were not satisfied in regard to the
generated code amount. In decoding, an ordinary algorithm of the
MPEG-2 which employs conventional inverse discrete cosine transform
was used. It is to be noted that the distance between core pictures
(I and/or P pictures) was M=3 (two B pictures are inserted between
I and/or P pictures), and the number of frames in a picture group
(GOP; Group of Pictures) was N=15.
[0089] Average values of PSNR (Peak Signal to Noise Ratio) values
of decoded pictures obtained in such a manner as described above
over 150 frames (120 frames for the picture "Confetti") are
illustrated in FIG. 6.
[0090] From FIG. 6, it can be seen that the reversible moving
picture coding systems of the first to third embodiments described
above exhibit augmented picture qualities comparing with that
obtained by the conventional system. The reversible moving picture
coding system of the second embodiment exhibits the highest
augmentation. In the reversible moving picture coding system of the
second embodiment, reversible discrete cosine transform is actually
performed for both of a frame-scanned block and a field-scanned
block, and that scanning which exhibits a higher augmentation in
decoded picture quality is selected. Consequently, while the
reversible moving picture coding system of the second embodiment
requires the largest amount of calculation among the reversible
moving picture coding systems of the first to third embodiments, it
exhibits the highest picture quality.
[0091] The reversible moving picture coding system of the third
embodiment exhibits the second highest decoded picture quality. In
the reversible moving picture coding system of the third
embodiment, in place of actually performing reversible discrete
cosine transform for both of a frame-scanned block and a
field-scanned block, a decoding predictive error is estimated by
approximation calculation. Since approximation calculation is
involved in this manner, although the reversible moving picture
coding system of the third embodiment exhibits a degree somewhat
inferior to that of the reversible moving picture coding system of
the second embodiment which is the best choice, it exhibits a
reduced calculation amount.
[0092] The reversible moving picture coding system of the first
embodiment exhibits the third best decoded picture quality. In the
reversible moving picture coding system of the first embodiment,
calculation is simplified by employing approximation up to the step
in which a predictive picture is produced in the reversible moving
picture coding system of the third embodiment. Consequently, while
the reversible moving picture coding system of the first embodiment
exhibits the lowest degree of augmentation among the reversible
moving picture coding systems of the first to third embodiments, it
involves the least amount of calculation among them.
[0093] While preferred embodiments of the present invention have
been described using specific terms, such description is for
illustrative purposes only, and it is to be understood that changes
and variations may be made without departing from the spirit or
scope of the following claims.
* * * * *