Efficient algorithms for MPEG video compression 1st Edition Dzung Tien Hoang download
Efficient algorithms for MPEG video compression 1st Edition Dzung Tien Hoang download
https://ebookgate.com/product/efficient-algorithms-for-mpeg-
video-compression-1st-edition-dzung-tien-hoang/
https://ebookgate.com/product/h-264-and-mpeg-4-video-compression-
video-coding-for-next-generation-multimedia-1st-edition-iain-e-
richardson/
ebookgate.com
https://ebookgate.com/product/video-compression-and-
communications-2nd-edition-hanzo/
ebookgate.com
https://ebookgate.com/product/digital-video-and-hdtv-algorithms-and-
interfaces-1st-edition-charles-poynton/
ebookgate.com
Compression for Multimedia 1st Edition Irina Bocharova
https://ebookgate.com/product/compression-for-multimedia-1st-edition-
irina-bocharova/
ebookgate.com
https://ebookgate.com/product/principles-of-filtration-1st-edition-
tien/
ebookgate.com
Contents
Preface xv
Acknowledgments xix
Acronyms xxi
1 Preliminaries 1
1.1 Digital Video Representation 1
1.1.1 Color Representation 2
1.1.2 Digitization 3
1.1.3 Spatial Sampling 4
1.1.4 Temporal Sampling 4
1.1.5 Quantization 5
1.1.6 Standard Video Data Formats 5
1.2 A Case for Video Compression 8
1.3 Spatial Redundancy 9
1.3.1 Vector Quantization 9
1.3.2 Block Transform 10
1.3.3 Discrete Cosine Transform 10
v
vi CONTENTS
xiii
Preface
In this chapter we survey aspects of video compression that will be useful for
understanding the later chapters of this book. After a brief discussion about
the digital representation of video, we motivate video compression with two
illustrative examples that underscore the need for lossy compression. We then
describe basic lossy compression techniques for reducing spatial and temporal
redundancy in video, focusing on techniques commonly used in international
standards for video coding. Next, we present an overview of some of the more
popular standards for video coding, with special focus on the bit rate control
algorithms used in the development and evaluation of the standards. We
conclude with an introduction to rate-distortion theory and the operational
rate-distortion framework that forms the basis of much work on optimal bit
rate control of compressed video.
This chapter is by no means intended to be comprehensive; for more in-
depth discussions of the fundamentals and applications of video coding, the
reader is referred to [2, 26, 55, 57, 67, 73].
1
2 PRELIMINARIES
1In the 17th century, Isaac Newton discovered that a small number of colors from the
spectrum produced by a prism can be mixed to produce other colors, including those not
in the spectrum.
DIGITAL VIDEO REPRESENTATION 3
red/green/blue
filter
Fig. 1.1 Block Diagram of a Video Digitizer. This gure shows the typical
processing steps involved in the digitization of video. After signal acquisition and
ampli cation, the key processing steps are spatial sampling, temporal sampling, and
quantization.
1.1.5 Quantization
After spatial and temporal sampling, the video signal consists of a sequence of
continuous intensity values. The continuous intensity values are incompatible
with digital processing, and one more step is needed before this information
can be processed digitally. The continuous intensity values are converted to
a discrete set of values in a process called quantization .
Quantization can be viewed as a mapping from a continuous domain to
a discrete range.2 A particular quantization mapping is called a quantizer.
An example is shown in Figure 1.3. In the gure, there are eleven discrete
quantization levels, also called bins. Each bin has an associated size, which
is the extent of the continuous values that map to that bin. In the example,
each bin, except for the bins for ,5, 0, and 5, has the same size, which is
sometimes referred to as the quantizer step size . This type of quantizer is
called a uniform quantizer. A binary encoding can be assigned to each of
the bins. Typically the initial quantization of a continuous source is done
using a number of quantization levels that is a power of 2, so that a xed
number of bits can be used to represent the quantized value.3 This process
of representing a continuous value by a nite number of levels using a binary
code is often referred to as pulse code modulation (PCM).
In conclusion, after spatial sampling, temporal sampling, and quantization,
we have N M picture elements, commonly called pixels or pels , represented
using a xed number of bits.
2 This de nition is intended also to encompass mappings from a discrete domain to a discrete
range.
3 Further quantization of digitized data may use a number of quantization levels that is not
a power of 2 and employ variable-length entropy coding.
4 The CCIR has changed its name to the International Telecommunication Union Radio-
communication Assembly (ITU-R), and the latest revision of the CCIR-601 standard is
formally known as Recommendation ITU-R BT.601-5. We use the term CCIR-601 since it
is still in common usage.
6 PRELIMINARIES
2
Quantized value
-2
-4
-6
-8 -7 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7 8
Continuous value
Fig. 1.3 Example of a Uniform Quantizer. This example shows a quantizer that
has 11 possible discrete output values, or bins.
digital encodings. In the 8-bit encoding, assuming that the RGB components
have been digitized to the range [0; 255], the YCrCb components are de ned
as follows [26]:
Y = 0:257R + 0:504G + 0:098B + 16;
Cr = 0:439R , 0:368G + ,0:071B + 128;
Cb = ,0:148R , 0:291G + 0:439B + 128:
The CCIR-601 standard de nes a family of digital video formats. The
most commonly used member of the family is the 4:2:2, 13.5 MHz format.
In this format, the luminance component is sampled at a rate of 13.5 MHz
with 720 active samples per line. The chrominance components, Cr and Cb ,
each are sampled at 6.75 MHz with 360 active samples per line. For NTSC,
this sampling yields 486 active lines per frame at 60 elds/sec. For PAL, the
sampling yields 576 active lines per frame at 50 elds/sec.
In terms of pixels per frame, the 4:2:2 CCIR-601 format speci es spatial
sampling of 720 486 for NTSC and 720 576 for PAL. Temporal sampling
is interlaced 60 elds/sec for NTSC and interlaced 50 elds/sec for PAL.
The chrominance components are subsampled horizontally with respect to
the luminance component to take advantage of the human visual system's
reduced spatial sensitivity to color. This subsampling process is referred to
as the 4:2:2 format and is depicted in Figure 1.4.
Source Input Format. The Source Input Format (SIF) speci es spatial sam-
pling of 360 240 and progressive temporal sampling at 30 frames/sec for
DIGITAL VIDEO REPRESENTATION 7
Fig. 1.4 4:2:2 Color Subsampling. With 4:2:2 chroma subsampling, the two
chroma components are subsampled by a factor of two horizontally. The positioning
of the chrominance values relative to the luminance values is shown as speci ed by the
CCIR-601 standard.
5 For some applications, such as MPEG-1 and MPEG-2 video, it is convenient for the spatial
dimensions to be a multiple of 16. For this reason, a horizontal dimension of 352 is often
used.
6 When the horizontal image dimension of 352 is used, the horizontal chrominance sampling
would correspondingly be 176.
7 The 4:2:0 format should not be confused with the 4:1:1 format in which the chrominance
components are subsampled by a factor of 4 only in the horizontal direction.
8 PRELIMINARIES
Fig. 1.5 4:2:0 Color Subsampling. With 4:2:0 subsampling, the chroma compo-
nents are subsampled by a factor of two both horizontally and vertically. The position-
ing of the chrominance values relative to the luminance values is shown as speci ed
in the MPEG-2 standard. In the MPEG-1 standard, the chrominance samples are
positioned in the center of four adjacent luminance samples.
To transmit video in this format over a 28.8 Kbits/sec modem would require a
compression ratio of about 106:1. At such a high compression ratio, depending
upon the complexity of the video sequence, the quality of the compressed video
may have to be sacri ced. Alternatively, the frame rate could be reduced to
increase the image quality, at the expense of increased jerkiness in the motion.
The above examples show why compression is a must for some important
digital video applications. For example, without compression, a single-sided
DVD can hold less than four minutes of CCIR-601 digital video.
i=0 j =0
where F (u; v) are the transform coecients and
8
< 1
C (x) = : p2 if x = 0;
1 otherwise.
SPATIAL REDUNDANCY 11
2 3
8 16 19 22 26 27 29 34
6
6 16 16 22 24 27 29 34 37 7
7
6
6 19 22 26 27 29 34 34 38 7
7
6
6 22 22 26 27 29 34 37 40 7
7
6
6 22 26 27 29 32 35 40 48 7
7
6
6 26 27 29 32 35 40 48 58 7
7
4 26 27 29 34 38 46 56 69 5
27 29 35 38 46 56 69 83
Fig. 1.6 Default Intraframe Quantization Matrix. This gure shows the de-
fault MPEG-2 intraframe quantization matrix to be applied to 2D-DCT coecients.
With this quantization matrix, transform coecients are quantized more coarsely with
increasing horizontal and vertical spatial frequencies.
u=0 v=0
Quantization. Since the DCT and IDCT are transform pairs, they do not
result in any compression by themselves. Compression is achieved by subse-
quent quantization of the transform coecients.
Quantization as applied to transform coecients can be viewed as division
followed by integer truncation. Speci cally, the transform coecients are rst
divided by a (prespeci ed) matrix of integers that is weighted by a quantiza-
tion scale Q. After division, the results are truncated to integer values. In
the dequantization, the quantized values are multiplied by the quantization
matrix and adjusted according to the quantization scale. Typically 8 to 12
bits of precision are used.
An example of a quantization matrix is shown in Figure 1.6. The coe-
cients can be speci ed to exploit properties of the human visual system. Since
the human eye is more sensitive to low spatial frequencies and less sensitive
to high spatial frequencies, the transform coecients corresponding to high
spatial frequencies can be quantized more coarsely than those for low spatial
frequencies. This frequency-selective quantization can be seen in Figure 1.6.
Zig-Zag Scan. Because of the coarse quantization of coecients correspond-
ing to high spatial frequencies, those coecients are often quantized to 0.
An e ective way to code the resulting set of quantized coecients is with
a combination of a zig-zag scan of the coecients as shown in Figure 1.7
and run-length encoding of consecutive zeros. Typically, the DC coecient,
F (0; 0), is coded separately from the other coecients and is not included in
the zig-zag scan.
12 PRELIMINARIES
Horizontal
Frequency
0 1 2 3 4 5 6 7
DC
0
Frequency
Vertical
3
Fig. 1.7 Zig-Zag Scan. This gure shows a zig-zag scanning pattern for coding
quantized transform coecients as a one-dimensional sequence. A run-length encoding
of the zero values of the one-dimensional sequence is then performed.
Encoder
Frame
Frame Buffer
Decoder
Fig. 1.8 A Simple Frame-Di erencing Coder. This block diagram shows a
simple frame-di erencing coder. The frame bu er stores the previously decoded frame
which is used to compute a di erence frame.
ing fails is to switch to intracoding mode. However, this technique does not
improve the compression eciency of frame di erencing.
Encoder
Motion information
Motion Compensation
are reordered so that all reference frames that are required by a B frame or P
frame come before that frame in the reordered sequence. An example is shown
in Figure 1.13. In practice, this reordering introduces some encoding and de-
coding delays and requires two frame bu ers to hold the reference frames.
For non-real-time applications, such as stored video, the additional delay is
not a serious issue. For real-time applications, such as video conferencing,
the distance between successive reference frames are kept small to reduce the
delay. B frames may be omitted altogether to further reduce the delay.
For interframe coding, perceptual weighting as per Figure 1.6 is not usually
applied since the block to be coded is the block of prediction errors, which
does not share the perceptual properties of the original spatial block of pixel
values. The MPEG standards specify a default interframe quantization ma-
trix with uniform values, as shown in Figure 1.11. In practice, the modi ed
interframe quantization matrix shown in Figure 1.12 is often used. In this
matrix, the quantization step size increases gradually as the frequency index
increases. This modi ed interframe quantization matrix works well in prac-
tice and has been adopted as the default matrix in the MPEG-4 visual coding
standard [36].
In the nal step, indicated in Figure 1.9, the prediction error that results
from motion compensation is coded with an intraframe coder, for instance,
one of the techniques mentioned in Section 1.3.
I B B P B B P
Fig. 1.10 Frame Types for Motion Compensation. This gure illustrates three
types of frames use in motion compensation. An I frame is coded using intraframe
techniques and has no temporal dependencies. A P frame is predicted from a previous
reference frame, which may be either an I frame or a P frame. B frames are predicted
from a past and a future reference frame.
2 3
16 16 16 16 16 16 16 16
6
6 16 16 16 16 16 16 16 16 7
7
6
6 16 16 16 16 16 16 16 16 7
7
6
6 16 16 16 16 16 16 16 16 7
7
6
6 16 16 16 16 16 16 16 16 7
7
6
6 16 16 16 16 16 16 16 16 7
7
4 16 16 16 16 16 16 16 16 5
16 16 16 16 16 16 16 16
Fig. 1.11 Default Interframe Quantization Matrix. This gure shows the de-
fault MPEG-2 interframe quantization matrix to be applied to 2D-DCT coecients.
With this quantization matrix, transform coecients are quantized the same indepen-
dent of frequency.
16 PRELIMINARIES
2 3
16 17 18 19 20 21 22 23
6
6 17 18 19 20 21 22 23 24 7
7
6
6 18 19 20 21 22 23 24 25 7
7
6
6 19 20 21 22 23 24 26 27 7
7
6
6 20 21 22 23 24 26 27 28 7
7
6
6 21 22 23 24 26 27 28 30 7
7
4 22 23 24 26 27 28 30 31 5
23 24 25 27 28 30 31 33
Fig. 1.12 Modi ed Interframe Quantization Matrix. This gure shows a mod-
i ed MPEG-2 interframe quantization matrix that is often used in practice and has
been incorporated into the MPEG-4 standard as the default interframe matrix [36].
With this quantization matrix, transform coecients are quantized gradually coarser
as the frequency increases.
Frame Type: I B B P B B P B I
Temporal Index: 1 2 3 4 5 6 7 8 9
(a) Original Sequence (Temporal Order)
Frame Type: I P B B P B B I B
Temporal Index: 1 4 2 3 7 5 6 9 8
(b) Reordered Sequence (Encoding Order)
Fig. 1.13 Reordering with B Frames. This gure shows the reordering of I and
P frames to allow for causal decoding of B frames. A reference frame (I or P frame)
is moved immediately before the rst B frame that uses it for backward prediction.
H.261 STANDARD 17
Fig. 1.14 Illustration of the Block Translation Model. In this gure, two
blocks in the Current Frame are shown to be copied from a similarly sized region in
the Reference Frame. This action forms the basis of the block translation motion
model.
8The CCITT has since changed its name to the International Telecommunication Union
Telecommunication Standardization Sector (ITU-T).
18 PRELIMINARIES
Y Y
16 8 CR 8 CB
Y Y
8 8
16
Fig. 1.15 Structure of a Macroblock. This gure shows the structure of a mac-
roblock when the 4:2:0 chroma format is used. The macroblock consists of four adjacent
8 8 blocks of Y samples, a co-sited 8 8 block of Cr samples, and a co-sited 8 8
block of Cb samples. This macroblock structure is used by H.261, H.263, MPEG-1,
MPEG-2, and MPEG-4 for 4:2:0 sources.
1.5.1 Features
The H.261 standard uses a combination of block matching motion compensa-
tion (BMMC) and 2D-DCT coding, as described in Sections 1.3.2 and 1.4.3.
Since H.261 is intended for real-time video conferencing applications, there is
a requirement for low encoding delay, which precludes the use of bidirectional
predictive motion compensation. Therefore only intraframe coding and for-
ward predictive coding are used, with a predicted block depending only upon
the previous frame. The real-time requirement also restricts the complexity
of higher-level algorithms, such as motion estimation and rate control.
The Common Intermediate Format (CIF) and Quarter-CIF (QCIF), de-
scribed in Section 1.1.6, are speci ed for video frames. A video frame is
divided into Groups of Blocks (GOBs) made up of a number of macroblocks
(MBs). As depicted in Figure 1.15, each macroblock is composed of four 8 8
luminance blocks and two 8 8 chrominance blocks, one each for the Cb and
Cr color components. Integer-pel motion compensation is performed at the
macroblock level; that is, there is one motion vector per macroblock.
-p
CC -t
-a
q? - qz
Video In q - m -q q - T - Q q
?
-q
6 Q,1
? To Video
Multiplex
T,1 Coder
a q - ?m
q -q ?
F P -v
6 -f
T: Transform p: Flag for INTRA/INTER
Q: Quantizer t: Flag for transmitted or not
P: Picture Memory with motion- qz: Quantizer indication
compensated variable delay q: Quantizing index for transform
F: Loop Filter coecients
CC: Coding Control v: Motion vector
f: Switching on/o of the loop lter
Fig. 1.16 Block Diagram of an H.261 Video Coder. This gures shows a block
diagram of a typical H.261 video coder [23].
160
128
Variance of original block
Interframe
motion compensation
96 y=x
64
Intraframe
32 motion compensation
0
0 32 64 96 128 160
Variance of motion compensated prediction error
5.5
5
MAD with estimated motion vector
4.5
4 Zero displacement
3.5 motion compensation y=x/1.1
3 2.7
2.5
2 Motion vector
1.5 compensation
1.5
1
0.5
0.5
0
0 1 2 3 4 5 6
MAD with zero motion vector
Fig. 1.18 Motion Vector Decision Diagram. This gure shows the motion vector
decision diagram of Reference Model 8 [4]. The mean absolute di erence (MAD) is
used as the measure of motion compensated prediction error. A motion-estimation
search is performed to determine a candidate motion vector with the least MAD. The
MAD with the candidate motion vector is compared with the MAD with the zero
motion vector to determine whether the candidate motion vector or the zero motion
vector is used for motion compensation.
Other documents randomly have
different content
He laughed a little. “Life has been too easy for me, Jim: I’m not used
to big decisions.”
“Well, I am a beast,” said Jim, frankly. “I’ve been chewing over
my own disappointment; and about the worst part of it was that I
got hold of the idea that you had put it right out of your mind, and
that you didn’t care. I wish I had known you were up to your eyes in
worry. But you never let us suspect a thing.”
“Well, I kept hoping against hope that each mail would straighten
things out,” his father answered. “Until I was certain I did not want
to cast any shadows on Norah’s holidays. Poor little lass; she’ll have
trouble in earnest now.”
“Well, Nor will face it,” Jim said, confidently. “She isn’t made of
the stuff that caves in—and as far as I’m concerned, Dad, she wants
me to go. She knew I’d only eat my heart out if I didn’t. But to have
you go away is another matter. Don’t you think you can take her?”
“If I were sure England would be safe . . .” mused Mr. Linton.
“You can be very certain I don’t want to leave her.”
“Well, I don’t think there’s much risk for England,” said Jim, with
the cheerful optimism of youth. “And anyhow, there’s always
America—you and she could slip across there if there were any real
fear of invasion. My word, Dad, it would be grand to think you and
Nor were so near. Just think if I got wounded, how jolly it would be
to come over to you!”
“I’ve thought,” said his father, drily. The jollity of the idea seemed
to him slightly exaggerated.
“Well, it would be heaps better than hospital. And then we’d all
be together after the finish, and do London. It would be such a lark.
Fancy old Norah in Piccadilly!”
“Me?” asked a startled voice.
Norah stood in the doorway, with Wally behind her. She had
exchanged her riding-habit for a soft white frock, and her brown
curls, released from their tight plait, fell softly round her face. No
one would have dreamed of calling her pretty; but there was an
indefinable charm in the merry face, lit by straight grey eyes. She
was tall for her age; people found it difficult to believe that she was
not yet sixteen, for she had left the awkward age behind her, and
there was unstudied grace in the slender, alert form, with its well-
shaped hands and feet. Occasionally—when she was not too busy—
Norah had fleeting moments of regret, mainly on account of her
men-folk, that she was not pretty. But it is doubtful if her father and
brother would have cared to change a feature of the vivid face.
“Did you say Piccadilly? And me?” she asked, advancing into a
startled silence. “I’ve always imagined Piccadilly must be rather
worse than Collins Street, and I don’t fit in there a bit. Stella
Harrison says there are rather jolly motor-busses there, and you can
get on top. That wouldn’t be so bad.” She perched on the arm of her
father’s chair. “Why are you talking about streets, Daddy? You know
you don’t like them any more than I do.”
“No,” said David Linton, finding that some answer was expected
of him. Something in his tone brought Norah’s eyes upon him
quickly.
“There’s something wrong, isn’t there?” she asked.
No one spoke for a moment. Then Wally got up quietly and
moved towards the door.
“Don’t go, Wally, my boy,” Mr. Linton said. “You’re so much one of
the family that you may as well join the family councils. No, there’s
nothing exactly wrong, Norah. But there are happenings.”
“Jim’s going?” said Norah, quickly. Her keen eyes saw that the
new and unfamiliar shadow had lifted from her brother’s face. Jim
nodded, smiling at her.
“Yes, I’m going. Dad says it’s all right.”
Norah drew a long breath, and Wally gave an irrepressible
whistle of delight.
“Lucky dog—I’m so glad!” he cried. “Oh, why can’t I be
eighteen!”
“There will be plenty of fighting after you are eighteen,” Mr.
Linton said. “This isn’t going to be any lightning business. But that’s
not all, Norah. Your old father has to pack up, too. I must go to
England.”
“Daddy! You!”
The voice was a cry. Then Norah shut her lips tightly, and said
nothing more, looking at her father.
“It’s business,” he said hurriedly. “I don’t want to go, my girl. It
may not take me long.”
There was a long pause.
“I can’t ask to go,” said Norah at last, rather breathlessly. “It’s too
big a thing—not like a trip to Melbourne or Sydney. I know it would
cost a fearful lot of money—and there are other things. It’s—it’s all
right, Daddy, if you say so—only I want to know. Have I got to stay
behind?”
There was no answer. Jim was watching the set, childish face
pitifully, longing to help, and powerless. Norah got up from the arm
of her father’s chair at length, and turned her face away.
“It’s—it’s quite all right, Daddy,” she said, unsteadily. “I
understand. Don’t go worrying.”
“Worrying!” said David Linton, explosively. “No, I’m not going to
worry—if I can help it: and I’m not going to leave you, either. We’ll
stick together, little mate.”
“Daddy!” said Norah, very low. She went to him like a little child,
and he put her on his knee, one arm round her, while Jim beamed
on them both.
“I knew you couldn’t do it,” he said laughing. “It was so
altogether ridiculous to think of old Nor here alone, and you and me
at the other side of the world. Things like that simply can’t occur!”
“Well—there may be danger” began his father.
“There would be strong danger of my losing my few wits if you
did it,” Norah said. “I thought I was going to lose them a minute
ago, as it was. Oh, Daddy won’t it be lovely! Think of the ship—and
the queer ports—and England! It’s the most wonderful thing that
ever happened. And we’ll be near Jim, and he’ll get leave and come
over to see us!”
“That’s another thing,” Mr. Linton said. “It’s settled that you’re to
enlist, Jim; that matter is decided. But is there any particular reason
why you should enlist in Australia?”
“In Australia?” repeated Jim, blankly. “Why—where else?”
“Well, if Norah and I are going home, why should we not all go
together? You would have no difficulty in joining the Army in
England, if boys of sixteen are getting commissions there.”
“What?” burst from Wally.
“Oh, yes—you’d be quite a veteran, judging by to-day’s news,
Wally,” said Mr. Linton, laughing. “There would be no difficulty at all,
I should think, Jim; I know enough people in London to pull a few
strings, though even that would hardly be necessary. But if you
wanted a commission I should think it could be managed. It would
leave us all together a bit longer.”
“That would be ripping,” Jim said, doubtfully. “I don’t know,
though; I’m an Australian, and I rather think Australians ought to
stick together. And I would know such a lot of the fellows in our own
contingent.”
“That counts, of course,” said his father. “But there’s another
point; there are rumours that our men may not be sent direct to the
Front. You might get hung up in Egypt, or the Persian Gulf, or Malta;
I’ve heard suggestions that the Australians should even be used for
garrison duty in India.”
“By Jove!” said Jim. “I wouldn’t like that.”
“No; and it would mean that you might never get to England at
all, to join Norah and me after the show. If you’re going, I don’t
want you to be shelved in some out-of-the-way corner of the earth;
I’d like you to have your chance.”
“Oh, Jimmy, come with us!” said Norah. “Just think how jolly it
would be—not like the voyage in a horrid old troopship, where you
mightn’t be allowed to see a single port. And perhaps we’d be
together quite a lot in England, before you were sent to the Front.”
Wally jumped up with such emphasis that his chair fell over
backwards. He did not notice it.
“Let’s all go!” he cried.
Three pairs of eyes turned upon him for information.
“If it’s really true that boys younger than I am are being taken in
England, I’d have a chance, wouldn’t I, Mr. Linton?”
“I suppose you would—yes, of course, my boy. You’re only a year
younger than Jim, aren’t you?”
“Yes—and he knows as much drill as I do, to say nothing of
shooting and riding,” Jim exclaimed. “Would you come, Wal?”
“I should just think I would!” Wally uttered. “But you’d have to
join in England, Jim—not here.”
“But your guardian—and your brothers, Wally. Would they be
willing?” Mr. Linton asked. “It’s rather an undertaking to arrange off-
hand. And it would mean your leaving school.”
“I know it would be all right, sir,” Wally answered. “My brothers
were only sorry I couldn’t get into the first contingent; and old Mr.
Dimsdale never worries his head about me, except to look after the
property and send me my allowance. He knows I’m to join as soon
as I can. The money part of it would be all right; I don’t know much
about it, but the money that’s to come to me has been accumulating
since I was a kid, and there must be plenty. If you’d let me go under
your wing, nobody would think of objecting.” He stopped, his brown,
eager face flushing. “By Jove, you must think me awfully cool, sir. I
sort of took it for granted I could go with you!”
“Well, you old goat!” said Jim, disgustedly. David Linton laughed.
“My dear boy, I think you’re pretty well established as one of the
family,” he said. “You have been Jim’s chum for five years, and
somehow we’ve come to regard Billabong as your home. I have liked
to think you felt that way about it, yourself.”
“It’s the only real home I ever remember,” said Wally, still greatly
confused. “And you’ve all been such bricks to me. I’ve quite
forgotten I’m really a sort of lost dog.”
“It’s rude to say you’re a lost dog, when you belong to Billabong,”
said Norah solemnly, though her eyes were dancing. “Isn’t he talking
a lot of nonsense, Dad?—and this is much too exciting an evening to
waste any time. I wish someone would sort me out, for I’m all
mixed-up in my mind. We’re going to England, you and I, Dad.”
“And me,” said Wally, cheerfully disregarding grammar.
“And me, I suppose,” Jim followed. “If you think I’ve as good a
chance there, Dad?”
“Better, I should think—judging from the rush of men here,” said
his father.
“Then we’re all going,” finished Norah blissfully. “In a ’normously
large ship, Dad?”
“Most certainly,” said David Linton, hastily. “I came out forty
years ago in a five-hundred tonner, and I’ve no desire to repeat the
experience. We’re built on lines that demand space, we Lintons.”
“And when we get to London?”
“We’ll settle down somewhere—where we can be near the boys
until they are sent out to the Front, and I can attend to business.”
“And then——?”
“We’ll wander about a bit until they come back to us. If it’s likely
to be long, you’ll have to resume your neglected education, young
woman,” said her father severely.
“M’f!” said Norah, wrinkling her nose. “How unpleasant!—that’s
the first dismal thing you’ve said, Daddy. But I suppose one has to
take the powder with the jam. And after the war——?”
“Oh, after the war——” said David Linton; and fell silent, looking
at his son.
“After the war,” said Wally, happily, “we’ll all meet in London, and
see the Kaiser led in triumph down Piccadilly. My own preference
leads me to hope that it will be on a donkey with his face towards
the tail of the ass, but I’m sadly afraid the world has grown too
civilised.”
“Well, you can’t call him and his crowd civilised, anyhow,” Jim
said.
“No. But we’ll have to be, I suppose, to show how nicely we were
brought up. Anyhow, after that we’ll explore all the things we’ve
always wanted to see—London, and Stonehenge, and the Dublin
Horse Show, and Killarney, and David Balfour’s country, and
heathery moors, and the Derby, and punts on the Thames, and the
Dartmoor ponies, and——” Wally’s extraordinary mixture left him
breathless, but the others took up the tale.
“And English lanes——”
“And ruins—truly ruins——!”
“And old castles——”
“And woods and hedges——”
“And real hunting country——”
“And real hunts——!”
“And trout-streams——”
“And Irish loughs——”
“And then,” said Norah, as the dinner-gong clashed out its
summons,—“then——”
“If we’ve any money left!” put in her father.
“Or even if we haven’t,” said Norah, and smiled at him—“we’ll go
back to Billabong!”
CHAPTER III.
OF A CHESTNUT BABY.
ebookgate.com