03 MK PPT End to End Data

7/21/2019 03 MK PPT End to End Data

1/57

1

Computer Networks: A Systems Approach, 5eLarry L. Peterson and Bruce S. Davie

03 End-to-End Data (Chapter 7)

Multimedia Data(Section 7.2)

can be covered to explain how

the data that is generated bymultimedia applications is

encoded and compressed.

Copyright 2010, Elsevier Inc. All rights Reserved


2/57

2

Chapter7

Problem

Fro the net!or"#s perspective$ app%ication

pro&ras send essa&es to each other. Each o' these essa&es is ust an uninterpreted

strin& o' ytes.

Fro the app%ication#s perspective$ ho!ever$these essa&es contain various "inds o' data

arrays of integers, video 'raes$ %ines o' te*t$

di&ita% ia&es$ and so on. +n other !ords$ these

ytes have eanin&.


3/57

3

Chapter7

Problem

,e no! consider the pro%e o' ho! est to

encode the di''erent "inds o' data thatapp%ication pro&ras !ant to e*chan&e into yte

strin&s.


4/57

Chapter7

Ch!pter "#tline

Presentation Forattin& u%tiedia Data


5/57

$

Chapter7

Present!tion %orm!tting

ne o' the ost coon trans'orations o'

net!or" data is 'ro the representation used y

the app%ication pro&ra into a 'or that is

suita%e 'or transission over a net!or" and vice

versa. /his trans'oration is typica%%y ca%%ed

presentation formatting.


6/57

&

Chapter7


/he sendin& pro&ra trans%ates the data it !antsto transit 'ro the representation it usesinterna%%y into a essa&e that can e transittedover the net!or" that is$ the data is encoded ina message.

n the receivin& side, the app%icationtrans%atesthis arrivin& essa&e into a representation that itcan then process that is$ the essa&e isdecoded. Encodin& is soeties ca%%edar&uentmarshalling, and decodin& issoeties ca%%ed unmarshalling. /histerino%o&y coes 'ro the 1PC !or%d.


7/57'

Chapter7


Presentation 'orattin& invo%ves

encodin& and decodin& app%ication

data


8/57(

Chapter7


Bi&-endian and %itt%e-endian yte order

'or the inte&er 32$77$372.


9/57)

Chapter7


/a*onoy Data /ypes

/he 'irst 4uestion is !hat data types the syste is &oin& to

support. +n &enera%$ !e can c%assi'y the types supported y

an ar&uent arsha%%in& echanis at three %eve%s.

5t the %o!est %eve%$ a arsha%%in& syste operates on soe set o' basetypes. Typically, the ase types inc%ude inte&ers$ '%oatin&-point nuers$

and characters. /he syste i&ht a%so support ordina% types and

oo%eans.

5t the ne*t %eve% are flat typesstructures and arrays.

5t the hi&hest %eve%$ the arsha%%in& syste i&ht have to dea% !ith

complex typesthose types that are ui%t usin& pointers +n suary$ dependin& on ho! cop%icated the type syste is$ the tas"

o' ar&uent arsha%%in& usua%%y invo%ves convertin& the ase types$

pac"in& the structures$ and %ineari6in& the cop%e* data structures$ a%%

to 'or a conti&uous essa&e that can e transitted over the net!or"


10/5710

Chapter7


5r&uent arsha%%in& convertin&$

pac"in&$ and %ineari6in&


11/5711

Chapter7


Conversion Strate&y nce the type syste is esta%ished$ the ne*t issue is

!hat conversion strate&y the ar&uent arsha%%er !i%%

use. /here are t!o &enera% options canonical intermediate form andreceiver-makes-right


12/5712

Chapter7


Conversion Strate&y /he idea o' canonica% interediate 'or is to sett%e on an

e*terna% representation 'or each type the sendin& host

trans%ates 'ro its interna% representation to this e*terna%

representation e'ore sendin& data$ and the receiver

trans%ates 'ro this e*terna% representation into its %oca%

representation !hen receivin& data

/he a%ternative$ !hich is soeties ca%%ed receiver-a"es-

ri&ht$ has the sender transit data in its o!n interna% 'orat

the sender does not convert the ase types$ ut usua%%y has

to pac" and '%atten ore cop%e* data structures. /he

receiver is then responsi%e 'or trans%atin& the data 'ro the

sender#s 'orat into its o!n %oca% 'orat


13/57

13

Chapter7


/a&s /he third issue in ar&uent arsha%%in& is ho! the

receiver "no!s !hat "ind o' data is contained in the

essa&e it receives. /here are t!o coon

approaches tagged anduntagged data. 5 ta& is any additiona% in'oration inc%uded in a essa&e8

eyond the concrete representation o' the ase types8that

he%ps the receiver decode the essa&e.

/he a%ternative$ o' course$ is not to use ta&s. 9o! does the

receiver "no! ho! to decode the data in this case: +t "no!s

ecause it !as pro&raed to "no!.


14/57

1

Chapter7


/a&s

5 3;-it inte&er encoded in a ta&&ed

essa&e

C


15/57

1$

Chapter7


Stus 5 stu is the piece o' code that ip%eents ar&uent

arsha%%in&.

Stus are typica%%y used to support 1PC. n the c%ient

side$ the stu arsha%%s the procedure ar&uents into

a essa&e that can e transitted y eans o' the

1PC protoco%.

n the server side$ the stu converts the essa&e

ac" into a set o' varia%es that can e used as

ar&uents to ca%% the reote procedure.

Stus can either e interpreted or copi%ed

C


16/57

1&

Chapter7


Stus

Stu copi%er ta"es inter'ace description as input

and outputs c%ient and server stus.

C


17/57

1'

Chapter7


E*ap%es


18/57

1(

Chapter7


E*ap%es


19/57

1)

Chapter7


E*ap%es


20/57

20

Chapter7


E*ap%es


21/57

21

Chapter7


E*ap%es [email protected]

5stract Synta* @otation ne ([email protected]) is an +S standard

that de'ines$ aon& other thin&s$ a representation 'or data

sent over a net!or".

/he representation-speci'ic part o' [email protected] is ca%%ed the BasicEncodin& 1u%es (BE1).

[email protected] supports the C type syste !ithout 'unction pointers$

de'ines a canonica% interediate 'or$ and uses type ta&s.

+ts stus can e either interpreted or copi%ed.

ne o' the c%ais to 'ae o' [email protected] BE1 is that it is used y

the +nternet standard Sip%e @et!or" ana&eent Protoco%

(S@P).

C


22/57

22

Chapter7



[email protected] represents each data ite !ith a trip%e o' the 'or

ta&$ %en&th$ va%ue

/he ta& is typica%%y an -it 'ie%d$ a%thou&h [email protected] a%%o!s 'or

the de'inition o' u%ti-yte ta&s. /he %en&th 'ie%d speci'ies ho! any ytes a"e up the va%ue

Copound data types$ such as structures$ can e

constructed y nestin& priitive types

C


23/57

23

Chapter7



Copound types created y eans o'

nestin& in [email protected]=BE1

[email protected]=BE1 representation 'or a 2-yte

inte&er

C


24/57

2

Chapter7


E*ap%es @D1

@et!or" Data 1epresentation (@D1) is the data-encodin&

standard used in the Distriuted Coputin& Environent

n%i"e


25/57

2$

Chapter7


E*ap%es @D1

[email protected]=BE1 representation 'or %en&th (a) A

yte () u%tiyte

@D1#s architecture ta&

C


26/57

2&

Chapter7


ar"up Lan&ua&es


27/57

2'

Chapter7


ar"up Lan&ua&es


28/57

2(

Chapter7


ar"up Lan&ua&es


29/57

2)

Chapter7


ar"up Lan&ua&es


30/57

30

Chapter7


ar"up Lan&ua&es


31/57

31

Chapter7


ar"up Lan&ua&es


32/57

32

hapter7


ar"up Lan&ua&es


33/57

33

hapter7

*#ltimedi! +!t!

u%tiedia data$ coprisin& audio$ video$ and sti%%

ia&es$ no! a"es up the aority o' tra''ic on the+nternet y any estiates. /his is a re%ative%y recent deve%opent8it ay e hard to

e%ieve no!$ ut there !as no ou/ue e'ore ;00J.

Part o' !hat has ade the !idespread transission o'u%tiedia across net!or"s possi%e is advances in

copression techno%o&y.

Because u%tiedia data is consued ost%y y

huans usin& their senses8vision and hearin&8andprocessed y the huan rain$ there are uni4ue

cha%%en&es to copressin& it.

Ch


34/57

3

hapter7

*#ltimedi! +!t!

ou !ant to try to "eep the in'oration that is ost

iportant to a huan$ !hi%e &ettin& rid o' anythin& thatdoesn#t iprove the huan#s perception o' the visua% or

auditory e*perience.

9ence$ oth coputer science and the study o' huan

perception coe into p%ay. +n this section !e#%% %oo" at soe o' the aor e''orts in

representin& and copressin& u%tiedia data.

Ch


35/57

3$

hapter7

*#ltimedi! +!t! /o &et a sense o' ho! iportant copression has een to the

spread o' net!or"ed u%tiedia$ consider the 'o%%o!in& e*ap%e. 5 hi&h-de'inition /M screen has soethin& %i"e A00 N AK;0 pi*e%s$

each o' !hich has ;2 its o' co%or in'oration$ so each 'rae is A00N AK;0 N ;2 G J0 and so i' you !ant to send ;2 'raes persecond$ that !ou%d e over AOps.

/hat#s a %ot ore than ost +nternet users can &et access to$ y a

&ood ar&in. By contrast$ odern copression techni4ues can &et a reasona%y

hi&h 4ua%ity 9D/M si&na% do!n to the ran&e o' A0 ps$ a t!o ordero' a&nitude reduction$ and !e%% !ithin the reach o' anyroadand users.

Sii%ar copression &ains app%y to %o!er 4ua%ity video such asou/ue c%ips8!e video cou%d never have reached its currentpopu%arity !ithout copression to a"e a%% those entertainin& videos'it !ithin the and!idth o' today#s net!or"s.

Ch


36/57

3&

hapter7

*#ltimedi! +!t!

Loss%ess Copression /echni4ues

+n any !ays$ copression is insepara%e 'ro data encodin&. /hat is$ in thin"in& aout ho! to encode a piece o' data in a set o' its$ !e

i&ht ust as !e%% thin" aout ho! to encode the data in the sa%%est set o'its possi%e.

For e*ap%e$ i' you have a %oc" o' data that is ade up o' the ; syo%s 5throu&h $ and i' a%% o' these syo%s have an e4ua% chance o' occurrin& inthe data %oc" you are encodin&$ then encodin& each syo% in J its is the

est you can do (since ;JG 3; is the %o!est po!er o' ; aove ;). +'$ ho!ever$ the syo% 1 occurs J0Q o' the tie$ then it !ou%d e a &ood

idea to use 'e!er its to encode the 1 than any o' the other syo%s.

+n &enera%$ i' you "no! the re%ative proai%ity that each syo%!i%% occur in the data$ then you can assi&n a di''erent nuer o'its to each possi%e syo% in a !ay that inii6es the nuer

o' its it ta"es to encode a &iven %oc" o' data. /his is the essentia% idea o' %uffman codes, one of the important

early developments in data compression.

Ch


37/57

3'

hapter7

*#ltimedi! +!t!

Loss%ess Copression /echni4ues 1un %en&th Encodin&

1un %en&th encodin& (1LE) is a copression techni4ue !ith

a rute-'orce sip%icity.

/he idea is to rep%ace consecutive occurrences o' a &iven

syo% !ith on%y one copy o' the syo%$ p%us a count o' ho!

any ties that syo% occurs8hence the nae >run

%en&th.?

For e*ap%e$ the strin& 555BBCDDDD !ou%d e encoded as

35;BAC2D.

Ch


38/57

3(

hapter7

*#ltimedi! +!t!

Loss%ess Copression /echni4ues Di''erentia% Pu%se Code odu%ation

5nother sip%e %oss%ess copression a%&orith is Di''erentia%

Pu%se Code odu%ation (DPC).

/he idea here is to 'irst output a re'erence syo% and then$

'or each syo% in the data$ to output the di''erence et!een

that syo% and the re'erence syo%.

For e*ap%e$ usin& syo% 5 as the re'erence syo%$ the

strin& 555BBCDDDD !ou%d e encoded as 5000AA;3333

since 5 is the sae as the re'erence syo%$ B has a

di''erence o' A 'ro the re'erence syo%$ and so on.

Ch

* lti di + t


39/57

3)

hapter7

*#ltimedi! +!t!

Loss%ess Copression /echni4ues Dictionary ased ethods

/he 'ina% %oss%ess copression ethod !e consider is the

dictionary-ased approach$ o' !hich the Lepe%-iv (L)

copression a%&orith is the est "no!n.

/he ni* copress and &6ip coands use variants o' the

L a%&orith.

/he idea o' a dictionary-ased copression a%&orith is to

ui%d a dictionary (ta%e) o' varia%e-%en&th strin&s (thin" o'

the as coon phrases) that you e*pect to 'ind in the data$

and then to rep%ace each o' these strin&s !hen it appears in

the data !ith the correspondin& inde* to the dictionary.

Ch

* lti di + t


40/57

0

hapter7

*#ltimedi! +!t!

Loss%ess Copression /echni4ues Dictionary ased ethods

For e*ap%e$ instead o' !or"in& !ith individua% characters in

te*t data$ you cou%d treat each !ord as a strin& and output

the inde* in the dictionary 'or that !ord.

/o 'urther e%aorate on this e*ap%e$ the !ord >copression?

has the inde* 2K7 in one particu%ar dictionary it is the

2K7th !ord in =usr=share=dict=!ords.

/o copress a ody o' te*t$ each tie the strin&

>copression? appears$ it !ou%d e rep%aced y 2K7.

Ch

* lti di + t


41/57

1

hapter7

*#ltimedi! +!t!

+a&e 1epresentation and Copression Oiven the increase in the use o' di&ita% ia&ery in

recent years8this use !as spa!ned y the invention

o' &raphica% disp%ays$ not hi&h-speed net!or"s8the

need 'or standard representation 'orats and

copression a%&oriths 'or di&ita% ia&ery data has&ro!n ore and ore critica%.

+n response to this need$ the +S de'ined a di&ita%

ia&e 'orat "no!n as IPEO$ naed a'ter the Ioint

Photo&raphic E*perts Oroup that desi&ned it. (/he>Ioint? in IPEO stands 'or a oint +S=+/ e''ort.)

Ch

* lti di + t


42/57

2

hapter7

*#ltimedi! +!t!

+a&e 1epresentation and Copression IPEO is the ost !ide%y used 'orat 'or sti%% ia&es

in use today.

5t the heart o' the de'inition o' the 'orat is a

copression a%&orith$ !hich !e descrie e%o!.

any techni4ues used in IPEO a%so appear in PEO$the set o' standards 'or video copression and

transission created y the Moving &icture 'xperts

(roup.

Ch

* lti di + t


43/57

3

hapter7

*#ltimedi! +!t!

+a&e 1epresentation and Copression Di&ita% ia&es are ade up o' pi*e%s (hence the

e&api*e%s 4uoted in di&ita% caera advertiseents).

Each pi*e% represents one %ocation in the t!o-

diensiona% &rid that a"es up the ia&e$ and 'or

co%or ia&es$ each pi*e% has soe nuerica% va%uerepresentin& a co%or.

/here are %ots o' !ays to represent co%ors$ re'erred to

as color spaces) the one ost peop%e are 'ai%iar !ith

is 1OB (red$ &reen$ %ue).

Ch

* lti di + t


44/57

hapter7

*#ltimedi! +!t!

+a&e 1epresentation and Copression ou can thin" o' co%or as ein& three diensiona%

4uantity8you can a"e any co%or out o' red$ &reenand %ue %i&ht in di''erent aounts.

+n a three-diensiona% space$ there are %ots o'di''erent$ va%id !ays to descrie a &iven point(consider Cartesian and po%ar co-ordinates$ 'ore*ap%e).

Sii%ar%y$ there are various !ays to descrie a co%orusin& three 4uantities$ and the ost coon

a%ternative to 1OB is M. /he is %uinance$rou&h%y the overa%% ri&htness o' the pi*e%$ and andM contain chroinance$ or co%or in'oration.

Ch

* lti di + t


45/57

$

hapter7

*#ltimedi! +!t!

+a&e 1epresentation and Copression Let#s %oo" at the e*ap%e o' the Oraphica% +nterchan&e

Forat (O+F). O+F uses the 1OB co%or space$ and starts out !ith its to

represent each o' the three diensions o' co%or 'or a tota% o';2 its.

1ather than sendin& those ;2 its per pi*e%$ ho!ever$ O+F'irst reduces ;2-it co%or ia&es to -it co%or ia&es.

/his is done y identi'yin& the co%ors used in the picture$ o'!hich there !i%% typica%%y e considera%y 'e!er than ;;2$ andthen pic"in& the ;J co%ors that ost c%ose%y appro*iate theco%ors used in the picture.

/here i&ht e ore than ;J co%ors$ ho!ever$ so the tric" isto try not to distort the co%or too uch y pic"in& ;J co%orssuch that no pi*e% has its co%or chan&ed too uch.

Ch

* lti di + t


46/57

&

apter7

*#ltimedi! +!t!

+a&e 1epresentation and Copression

B%oc" dia&ra o' IPEO copression

Ch

* lti di + t


47/57

'

apter7

*#ltimedi! +!t!

IPEO Copression

DC/ Phase DC/ is a trans'oration c%ose%y re%ated to the 'ast Fourier

trans'or (FF/). +t ta"es an

N atri* o' pi*e% va%ues as input and outputs an N

atri* o' 're4uency coe''icients.

ou can thin" o' the input atri* as a 2-point si&na% that is

de'ined in t!o

spatia% diensions (* and y) DC/ rea"s this si&na% into 2

spatia% 're4uencies.

Cha

* lti di + t


48/57

(

apter7

*#ltimedi! +!t!

IPEO Copression

DC/ Phase DC/$ a%on& !ith its inverse$ !hich is per'ored durin&

decopression$ is de'ined y the 'o%%o!in& 'oru%as

!here pi*e%(*$ y) is the &raysca%e va%ue o' the pi*e% at position

(*$ y) in the N %oc" ein& copressed @ G in this case

Cha

* ltimedi! +!t!


49/57

)

apter7

*#ltimedi! +!t!

IPEO Copression

Ruanti6ation Phase /he second phase o' IPEO is !here the copression

ecoes %ossy.

DC/ does not itse%' %ose in'oration it ust trans'ors the

ia&e into a 'or that a"es it easier to "no! !hat

in'oration to reove.

Ruanti6ation is easy to understand8it#s sip%y a atter o'

droppin& the insi&ni'icant its o' the 're4uency coe''icients.

Cha

*#ltimedi! +!t!


50/57

$0

apter7

*#ltimedi! +!t!

IPEO Copression

Ruanti6ation Phase /he asic 4uanti6ation e4uation is

Ruanti6edMa%ue(i$ ) G +nte&er1ound(DC/(i$ )=Ruantu(i$ ))

,here

Decopression is then sip%y de'ined as

DC/(i$ ) G Ruanti6edMa%ue(i$ ) N Ruantu(i$ )

Cha

*#ltimedi! +!t!


51/57

$1

apter7

*#ltimedi! +!t!

IPEO Copression

Encodin& Phase /he 'ina% phase o' IPEO encodes the 4uanti6ed 're4uency

coe''icients in a copact 'or.

/his resu%ts in additiona% copression$ ut this copression

is %oss%ess.

Startin& !ith the DC coe''icient in position (0$0)$ the

coe''icients are processed in the 6i&6a& se4uence.

5%on& this 6i&6a&$ a 'or o' run %en&th encodin& is used8

1LE is app%ied to on%y the 0 coe''icients$ !hich is si&ni'icant

ecause any o' the %ater coe''icients are 0. /he individua% coe''icient va%ues are then encoded usin& a

9u''an code.

Cha

*#ltimedi! +!t!


52/57

$2

apter7

*#ltimedi! +!t!

Mideo Copression (PEO)

,e no! turn our attention to the PEO 'orat$naed a'ter the ovin& Picture E*perts Oroup that

de'ined it.

/o a 'irst appro*iation$ a ovin& picture (i.e.$ video)

is sip%y a succession o' sti%% ia&es8a%so ca%%edframes or picturesdisplayed at somevideo rate.

Each o' these 'raes can e copressed usin& the

sae DC/-ased techni4ue used in IPEO.

Cha

*#ltimedi! +!t!


53/57

$3

apter7

*#ltimedi! +!t!


Se4uence o' +$ P$ and B 'raes&enerated y PEO.

Cha

*#ltimedi! +!t!


54/57

$

apter7

*#ltimedi! +!t!


Frae /ypes PEO ta"es a se4uence o' video 'raes as input and

copresses the into three types o' 'raes$ ca%%ed * frames

#intrapicture$, & frames #predicted picture$, and + frames

(idirectiona% predicted picture).

Each 'rae o' input is copressed into one o' these three

'rae types. + 'raes can e thou&ht o' as re'erence 'raes

they are se%'-contained$ dependin& on neither ear%ier 'raes

nor %ater 'raes.

Cha

*#ltimedi! +!t!


55/57

$$

apter7

*#ltimedi! +!t!


Each 'rae as a co%%ection o' acro%oc"s

Cha

*#ltimedi! +!t!


56/57

$&

apter7

*#ltimedi! +!t!


Forat o' an PEO-copressed video strea

Cha

#mm!ry


57/57

apter7

#mm!ry

,e have discussed ho! to represent ho! to

represent data in the net!or" ,e have discussed di''erent copression

techni4ues 'or hand%in& u%tiedia data in the

net!or"

Documents

03 MK PPT End to End Data