1‘»
`
`is
`
`r.
`
`if Digital TelevisiOn Transmission)
`’
`Using BandWidth
`- Coimpr‘esgilon Tec'briiques '
`Hisash‘r kaneko’ and Tatsuo tshi‘guro
`
`.
`
`“as
`
`l
`
`_,lGlTAL transmission and signal processingtechr
`niques have long been giewed as promising and
`. powerful means to achieve efficient television transmis
`sicfn by combining bandwidth compression with‘digital
`‘ transmission systems. Recent progress in LSl and digital
`technologies have made complicated signal processing a
`technically-feasible reality and have led to reasonable
`hardware size and cest. Highvcapamty and lowcost
`MOS memories have made it possible to store an entire
`television picture frame. This in turn has led to progress
`"in digital television encoding through the development of
`interframe coding, by which the transmission bit ratecan
`be greatly reduc’ed.
`‘
`M
`At the same time, progress in digital transmission sysv
`terns using multiphase modulation has allowed the use of
`digital
`transmission formats in microwave links and
`satellites. These trends have generated enthusiastic:
`efforts to develop and use digital approaches in actual
`television networksg
`;
`.
`‘
`In addition, the transmission of video teleconferencing
`services is growing rapidly. Teleconferencing has long
`‘ been anticipated as an alternative to travel and many
`trials and evaluation tests have and are being conducted.
`> Visual Communications is a key to teleconferencing. Full
`motion‘video transmission is thought to be very helpful
`and useful in ac‘complishinginteractive Communications.
`However, transmitting a full television Signal to accom~
`plish video teleconferencing would be 'very expensive,
`requiring a thousand times wider bandwidth thana voice
`telephone channel. Bandwidth compression is, there-
`fore, requiredto provide economical teleconferencing
`1 systems with compression ratios of 1:10 or even less.
`Thus, the growth of broadcast television as well as
`teleconferencing services will rely more and more ori
`sophisticated video processing techniques. These have
`’galready produced reasonable cost/performance results
`
`PMCAPL02442622
`
`and have made digital television transr‘rissgon a present
`reality
`actual field applications.
`’
`
`WHY DIGITAL TELEVISION
`TRANSMISSXON
`
`-_
`
`Digital transmiSSion-is Consioe red to be advantageous
`from the channel capacity point of view. particularly
`when it employs an efficient encoding scheme [6],[7l.
`Exisfing television transmission links,pre generally ana-
`log FM. They carry a single network quality television
`signal per radio channel. Although one might double the
`channel capaczty by restricting the modulation band-
`width {8] or by field interpolation muitiplex [9], the
`resultant tranSmission quality would not meet network
`television standards. Furthermore, with analog tech
`niques, highcompressionratios such as" 1: 10 could never
`a
`be achieved.
`'
`g
`fig.
`1 shows the relationship between-transmitter .
`power arid RF bandwidth of a‘radio link. For analog FM,~'
`the transmitter power
`increases significantly as RF
`bandwidth is reduced (at an SNR of 54 dB weighted). On
`"the other hand, digital transmission using quadraphase
`LF‘SK modulation. requires smaller RF bandwidth and
`smaller trandmitter power as the transmission bit rate
`reduces. At equivalent picture quality, digital technology
`offers obvious advantages over analog methodsat the bit
`rate below 60 Mbits/s.
`’
`.
`Furthermore, it carries with it various advantageous
`teatures of digital transmission. Picture quality is almost
`solely determined by the terminal equipment and can‘be
`made almost
`independent of transmission line impair-
`ments including those of digital terrestrial links. This
`implies
`that equal quality television service can be
`achieved uniformly over a wide service network.
`irre—
`spective of distance. Sophisticated digital processings
`can be employed SUCll‘lll'dl error control techniques
`improve the tolerance of the system to bit error rate,
`digital television signals can be easily.' enci'vpted to pro
`tect Comiiiunicaiimi primary, and adaptive hit sharing
`
`'
`
`The authors are with the Nippon Electric Company Ltd, Kawasaki.
`Japan.
`’
`‘
`’
`{1.
`3
`
`.114:
`
`. a
`
`IEEE Communications Magazine
`
`U]6368i)4/80/0700—0014 $00.75 @9 1980 [EEE
`
`PMC3683129
`
`PMC Exhibit 2025
`Apple v. PMC
`IPR2016-00755
`Page 1
`
`

`

`
`
`PRESENT FRME
`
`PREVIOUS FRME
`
`fRAHE DIVERENCE
`
`ANALOG FM
`SNR 54 dB
`
`Ilh LINE
`
`mvtm OBJECT
`
`;
`
`E
`
`a F
`
`NM LINE
`IAVEFGRH
`
`Fig. 3. Principle of basic Inlertrame coding.
`
`025nm
`a-vn var
`,
`75
`60 HthS/S
`
`ij
`
`45 MbitS/S
`
`3D Mbtts/s
`
`
`
`
`
`TRANSMITTERPOHER(dB)
`
`am
`Our
`o
`
`FEEDBACK CUNTROL
`7'
`forum ~
`
`A
`VARIABLE
`LENGTH
`BUFFER
`MEMORV
`CQDING .,
`
`
`
`
`
`lNTERFRAME
`50mm,
`
`achievable by use of interframe coding, in which the dif-
`ference signal between two successive frames is en-
`coded and transmitted. The general concept of bit rate
`reduction by interframe coding can hie better illustrated
`by a simplified model as shown in Fig 3. Signose that a
`soccer ball is crossing a televi5ion screen On looking at
`two successive telemsmn frames, the soccer ball in the
`present frame is positioned slightly differently from that -
`of the previous trainer and the difference corresponds to
`the movement of the soccer ball during the time period:
`One frame. Instead of transmitting the entire information,
`if the difference information of the two successrve frames
`is transmitted, thmamount of information to be trans
`mitted can be greatly reduced ln fact, if the movement is
`zero, tneoretically, no information need be transmitted
`The information to be transmitted is dependent on
`picture object movement:
`the more active the move-
`ment,
`the greater the information to be transmitted
`becomes.
`A
`The basic configuration of the interframe coder is
`shown in Fig. 4. The dignified televtsion Signal, converted
`by an A-to-D converter.
`is encoded by an interframe
`coder, in which essentially the difference signal between
`the present and the"previous frame is encoded. The
`previous frame signal is obtained from frame storage
`built in the interframe coder. The output of the encoder
`islagain processed through a variable length coder to
`remove the redundancy contained in the frame differen-
`tial Signal As mentioned earlier. the information rate is
`directly dependent on the movement of picture Obiects.
`lt should be smoothed out to obtain a constant trans.
`mission bit rate This function is accomplished through
`bufferJnemory aiid‘feedback control to the inierframe
`coder to suppress the excessive generation of informa
`tion. Since the control of the information generation rate
`is made by changing the quantizing step size. the signal-
`3
`
`PMCAPL02442623
`
`15 Mbits/s
`
`-lO
`
`0
`
`IO
`
`20
`
`(MHZ)
`RF BANDHIDTH
`,.
`Fig. 1: A compulson of analog FM and digital transmission in
`terms of RF bandwidth and transmitter.
`
`techniques can further reduce the required bit rate by
`sharing some of the total bit rate among plural channels.
`
`WHAT IS INTERFRAME CODING?
`
`Many investigations‘have been made to realize effi
`cient digital coding [111,[12]. Digital television encoding
`schemes are generally categorized into three classes:
`1) conventional PCM, 2) intraframe coding and 3) inter-
`frame coding, as shown in‘Fig. 2. For NTSC color tele-
`vision signals, conventional PCM or straight A~to~
`con
`version provides high-quality encoding‘ with 7- or 8-bit
`encoding at about a 10 MHz sampling, resulting in a 75
`through 86 Mbit/s transmi55ion rate [10].
`intraframe
`coding is- well known as a technique to reduce the trans
`mission bit rate by intraframe processing of the signal
`such as higher order differential PCM (HO—DPCM') or
`orthogonal transform coding. By these intraframe cod
`irig methods, the transmission bit rate can be reduced
`to about 30 through 60 Mbits/s depending on quality
`requirements and technique employed. There are also
`certain tradeoffs between picture quality, bit rate, and
`hardware complexity.
`Much greater reduction in transmission hit rate can be
`
`
`((Tel econ ference)
`'
`{ Network)
`I
`i\\‘
`1 :uniiis ‘
`NtERrRAMEDO
`
`:
`
`
`
`PICTUREQUALITY
`
`60
`40
`r.
`CDDfMG BIT RATE (Nblts/S)
`Fig. 2. Coding blt vale versus plolure quality for lyplcul coding
`methods
`
`80
`
`Jply 1980
`
`‘
`
`Hg 4. A basic contigurallon o! lntartrame encoder.
`
`PMC3683130
`
`PMC Exhibit 2025
`Apple v. PMC
`IPR2016-00755
`Page 2
`
`

`

`8 BIT/SIMPLE
`
`cw“
`gtun-virtues:
`r’ we
`PROCESS.§ mom:
`
`FEEDBACK
`2
`CONTROL
`(gqgt
`r"..._...m, ,V,__
`
`l Martian} r
`i—u
`‘
`iiier 'ii m
`: $019434?
`
`VIDEO
`1‘
`‘
`sown
`
`79;?
`SOUND
`
`“9
`
`com
`
`- 3 BITS/SAMPLE
`{AVERAGE}
`.
`
`”
`
`‘ BUFFER
`iFEMQRY‘
`
`ERROR
`iggpsa
`
`DArA
`OUT
`
`to-noise ratio is decreased as the amount of motion
`increases, When the information generation rate 8%
`ceeds the transmission rate, picture quality starts to
`degrade. The encoded picture V quality thus varies
`according to'the video source materials.
`As will be stated later, for network television signaés
`where much more active motion is encountered, ettceir
`lent picture quality can be transmitted at a bit rate
`around 20‘through 30 Mhits/s l24],[29]. Relatively still
`9‘ pictures such as those encountered in conference room
`: "scenes can be transmitted at 6 Mbits/s, 3 Moira/"s, or
`even lower bit rates [13]—[22].
`It should be noted here that the picture quaiity mea-
`sure for interframe Coding is different from the existing
`analog evaluation measure. Even at 3 Mbitsjs, the signal»
`to-noise performance for a stili picture is as good as that
`of 60 Mbitfis PCM. Picture quality impairments occur
`only when the picture moves actively. and the type of
`impairment
`is very much different from simpie noise
`impairment. Presently, no objective evaluation standard
`has been established. Picture quality canionlv be sub
`jectively evaluated.
`
`PMCAPL02442624
`
`our:
`
`UECGDER WW-W“..WWWW
`
`VXDEO
`Qt}
`
`an
`
`.
`i
`i
`.r
`P03?
`z rmrsermnt
`“4 Demetri:
`{PRIESSJ i
`
`‘
`
`Fig. a. An ifllfliflm encoder and decoder mock diagnm.
`
`iNTERFRAME CODING FOR NETWORK
`QUALITY TELEVISION
`The ei‘coders’clecocler arrangements for'high-quality
`transmission of broadcast quality is shown in Fig. 5
`{24],{29}. An input composite NTSC color television
`sigma.1 is first converted by an Ari—37D converter into a
`‘ 8 bit PCM signal. The PCM signal is then encoded into a
`reduced bit rate format of 2-3 iris/sample on average
`through digital signal processing in the network tele-
`vision appiieation, the transmission system is required
`to be transparent to the composite NTSC color video
`Signal; Therefore,
`the codec is designed so that the
`composite signal is directiy encoded and the waveform of
`the input signal
`is preserved except for quantization
`error. I
`a
`‘
`in an NiirSC color television szgnalg the coior subcar:
`rier phase is different by 1803 between successive
`frames. This is undes:rabie because it generates a large
`frame difference even when the picture is perfectly still.
`To solve this, phase inversion is made by a preprocess-
`ing circun in which'the composite signal is separated into
`luminance and chrominance components by reversible
`
`3
`
`
`
`
`
`IK‘OWTIONRATE(fibits/S)
`
`
`60
`
`Fly. 5. Variation at encoded data rule wlm Rims.
`u;
`
`lamt (s)
`
`alarm! source It "Superbowl ’79.“ most ncllve picture example.
`
`IELE Communicatitins Magazine
`
`PMC3683131
`
`PMC Exhibit 2025
`Apple v. PMC
`IPR2016-00755
`Page 3
`
`

`

`'PROBABILITY(it)
`
`7
`
`[TING ilME
`AVERAGE
`
`L_-... ;.A_.;_
`l3
`2')
`
`A.
`"ill
`
`..,,u¥4___
`bl]
`
`
`
`transformation and the chrominance component polar- ’
`ity is inverted frame by frame. After phase correction
`the signal
`is encoded by an interframe predictive
`encoder. Adaptive intrafram’e/interframe coding is used
`by selecting either an intraframe prediction or an inter-
`frame prediction, whichever is appropriate. For quantiv
`zation of
`the prediction error signal, eight
`ltinds oi
`qtihntization characteristics, QO-Q7, are prepared and
`onecb'f them is adaptively chosen through feedback con»
`trol depending on buffer memory occupancy value.
`Quantization steps for the small amplitude region are,
`1(QQ), 1.5(Q1), 2.5(Q3), 4.5(Q4), etc.
`respectively,
`where the magnitude 1 corresponds to one quantization
`step size of the 8 bit A/D converSion. The quantized
`prediction error is then coded with a variable length
`coder to reduce a bit rate. The compressed data are
`Stored in a buffer memory with a capacity of about 1
`Mbit, and is then transmitted to a line. Butter memory
`occupancy value (BMO) is fed back to the adaptive
`quantizer to control the encoded data generation rate in
`order to prevent buffer overflow.
`‘
`The performance of the interframe coder is better
`illustrated by way of example in Fig. 6. The example is
`taken from “Superbowl 79,” a most violent pictureexam
`pie, for a period of 2 min. The waveform shows the inter
`mation generated from the 'interframe coder without
`feedback control. The amount of information here indi-
`cates more or less original source information. As is
`seen,
`the amount of information. varies tremendously
`according to the source scenes. Then, by adding _
`feedback control, the system operates to allow coarse
`quantization for the portions of excessive source infor~
`mation, and as a result, keeps the output bit rate con-
`stant. This is the case for active motion, although gener-
`ally television programs yield less source information.
`Fig. 7 shows the statistical data of source information
`for various source scenes. The dotted line shows the
`probability of source information rate for various
`source materials taken from broadcast television pro-
`grams for 36 h [29]. It
`is seen that the average rate is
`around 15 Mbits/s, and as is seen on the solid line for
`the cumulative probability, 93 percent of the scenes
`are handled at 20 Mbits/s and 99 percent at 30
`Mbits/s without appreciable noise impairments.
`
`PMCAPL02442625
`
`AMOUNT OF INFORMATION Waits/s)
`
`Fig. 7.
`
`Statistics at the encoded data rate tor actual broadcast
`television slgnals tar as h In total.
`
`FURTHER IMPROVEMENT IS ACHIEVED BY
`ADAPTIVE BIT SHARING
`4
`THROUGH PLURAL CHANNELS
`‘5
`
`From what we have observed through these statistics,
`excellent picture quality can be obtained by-20 Mbitgs
`coding most of the time. Buffer till occurs with very small
`probability. According todficma‘l measurements, buffer
`fill=probability decreases \Zéry rapidly as the transmission
`bit rate is increased. The concept of sharing the trans
`mission bit rate among plural televiSion channels be
`comes quite effective in reducing the probabilities: of
`buffer fill, and in improving picture quality for extremely
`active motion.
`,
`'
`.
`Adaptive bit sharing is a—concept quite similar to that
`of TASI (time assignment speech interpolation) [27],
`using the advantage of statistical difference among plural
`channels. When a channel
`is transmitting a rapidly
`moving picture, other channels may be transmitting rela-
`tively quiet pictures, because the probability of occur-
`rence of rapid picture motion is generally very small.
`Therefore, we can assign a larger bit rate to the rapidly
`moving channel. and a smaller bit rate to the other rela-
`tively quiet channels, keeping the total bit rate constant.
`In fact,
`for
`three channel transmission with total 60
`.\\
`
`’
`
`SATELLITE
`
`UP LINK
`
`NETEC_22H >17
`ENCODER
`
`‘ 27 Hbt‘ts/s
`
`0mm LINK
`
`4! \
`/ \
`
`NETEC -22H 4—
`DECODER
`
`NETEC-ZZH
`ENCODER
`
`NETEC 722H
`ENCODIR
`
`Fly. 3. A three channel televlllon transmlsllon Iyltem.
`
`July 1980
`
`Mitt—“22H e-
`PULPL‘E’L
`
`PMC3683132
`
`PMC Exhibit 2025
`Apple v. PMC
`IPR2016-00755
`Page 4
`
`

`

`?_
`
`. Bi? RATE
`zc Mons/s.
`
`'
`
`Vlde
`ERROR COQRECUON
`
`
`
`MEAN{ERRORWEETHE(5)
`
`\ .
`,wirnotn
`\
`,
`i ERROR common
`
`CHANNEL BIT ERROR FATE
`
`Fig. 9. Error correction performance of double arm: mramicn
`(DEC) BCH code. The wild llne is theoretical and the circles
`are the Held test results.
`
`Mbit/s rate, the bit rate lor each channel can be adao
`tively varied within the 17 to 27 Mbit/s range, with an
`average of 20 Mbits/s.
`.
`‘
`lntertrame coding performance improvement by
`adaptive bit sharing is measured {29?
`:23an a real
`hardware system shown in Fig. 8. Three different broad-'7
`cast television program signals are supplied to the three
`encoders, and the adaptive bit rate assignment is per-
`formed ‘byan ABS-MUX (adaptive on sharing multio
`plexer) with the total bit rate kept constant at 60
`Mbits/s.
`
`Fig.
`
`intertrame encoderldecoddr
`the
`10. Photograph oi
`"'
`stereo-22H and ABS-MUX system.
`1,
`
`INTERFRAME CODTNG FOR VIDEO
`TELECONFERENCIfiG
`“ So much for the interfrarne encoding for high-quality”.
`izelwori‘t leievssioti. Now we wzll describe the inlerlrame
`technology for teleconferencing applications where the
`advantage oi anterframe coding is more significant
`in
`reducing the transmission bit rate down to '6 Mbitfi/S,
`3 Mints/s, or less. Basic principles of interframe coding
`are the same as those for network television, but there ,'
`extst differences in technologies related to achieve high
`data compression under conditions of slower picture
`movement. For this application, transparency is not as
`important as, :n broadcast network use. Algorithm design
`is therefore directed to a best compromise of the trans.»
`rnission bit rate and picture/quality.
`A functional block diagram of the interframe encoder
`and decoder terminal is shown in Fig. 11 l17l,[l8]. The
`terminal encodes color video and audio signals into a
`digital stream and multiplexes them into a bit stream 053
`(through 6 Mbiis,»s. Video signal input and output are
`NTSCacolor teevlsion signals‘whichore widely used in
`conventional video equipment, An audio signal with 7
`kHz bandwidth is encoded at a bit rate of 128 Rolls/s
`
`NTSC/TW
`
`nirmrizAiielm ri [JET—e
`LNCGCER
`i XIHTERWCE‘”
`‘-
`.
`,
`,
`ll
`i," ,,,,,,,,A__.m_,._.,l
`i.an Minis/s
`
`Auuio v~
`/
`,7,
`7,1
`A D
`\
`fl MT’Le,iim,
`
`“T
`T
`imsrmwl
`
`“l
`
`li'il {CSP‘RFEECODEF
`itil‘lfltflfiii
`~~ w”
`
`, ..
`*
`
`iii
`
`PMCAPL02442626
`
`,m
`'
`ERROR CONTROL
`This system has a forward acting error control circuit
`with 239/255 double error correcting BCH code, This
`error control has excellent error correction capability as
`shown in Fig. 9 [28]. The calculated mean error free time
`is longer than 1 h at a line bit error rate of 18‘s, and is
`about 5 5 even at 10“ which is generally considered the
`digital link threshold._The actual observed mean error
`free timé, obtained from a satellite transmission experi-
`ment, is plotted in the figure. These data coincide very
`well with calculated error control performance.
`ln
`ngrmal satellite conditions,
`the bit error rate gerton
`mance is generally better than 104, and the mean error
`free time gets to be much longer than at year.
`in the
`in addition to forward acting error control,
`A decoder, an erroneous line is automatically replaced by
`:he previous line, making the error less observable to the
`human eye.
`A typical realization oi‘ihe interframe encoders/decoder
`for network television ill the NE‘lVEC-‘BZH developed by
`Nippon Electric Company. Ltd. as shown in Fig, 10. This
`(.oritiguratron consists of three encoders (from the loll).
`an ABS Ml lX/l)l lX boy, a PSK MODEM> and two
`receiving decoders. Also each encoder/decoder con
`faint; two lb kHz oudlf! channels.
`
`T i
`VlDEO
`
`[JAIA
`
`an;
`“"‘
`
`Fig. 11 A digital television Iransmlsslon larmlnnl. NETEC—e/S.
`to: Inleconlemnclng use.
`
`“3H; Communications Magazine
`
`PMC3683133
`
`PMC Exhibit 2025
`Apple v. PMC
`IPR2016-00755
`Page 5
`
`

`

`ln order to accommodate color information, the com—
`posite NTSC signal
`is converted by a color signal
`processor, CSP-S, into a time division multiplex (TDM)
`signal which is a baseband video signal with a time comv
`pressed chrominance signal inserted into the horizontal
`blanking interval as shown in Fig. 12. (The interframe
`predictive coding used is an element-to-eiement differ-
`ence of frame-torlrame ditterence coding {191131}. The
`coding parameter/mode-controlgo prevent buffer over-
`flow in the codec is made by adaptive quantization. field
`repeating, subsampling, and frame freezing.
`The picture quality ‘of the interframe coding is de-
`scribed in relation to the coding parameter/mode-
`control. When buffer occupancy is in the lowest level, the
`encoding is made with the finest quantization step size to
`provide the highest signaltoquantization noise ratio.
`As buffer occupancy increases, the quantization step
`size is made coarser. The data generation rate reduces
`as the step size increases. When occupancy is turther
`raised and exceeds a certain level, a field repeating mode
`is applied in which only the information of every other
`field, either odd or even, is transmitted The deleted
`field is interpolated {mm the. adjacent fields at the receiv-
`ing side. By the field repeat mode: the information gen-
`eration rate can be halved. For
`further occupancy
`increase, a subsampling mode is added to the field
`repeating. The subsampling is to transmit every other
`sample. If the bgfter becomes full in spite of the controls
`described above: the encoding is stopped until the buffer
`occupancy, decreases to a low level. Al'the receiving side,
`the same picture is repeated, until a new frame picture is
`received. These mode controls give rise to a tradeoff
`between the data rate and picture quality and cause
`impairments as shown in Fig. 13. As the amount of mo-
`tion increases, the picture quality gradually decreases.
`in 6 Mbit/s encoding, a teleconferencing scene with
`natural motion of attendees can be transmitted within
`
`PMCAPL02442627
`
`.a________—V._.————————e-¢
`QUANTIZATION
`4‘
`FlNE
`COARSE
`
`“Aggééém” "
`
`(mot: comm)
`
`'
`
`, SLIGHT atzmtss ’
`r. RESOLUTZON LOSS
`BLLRRINB
`JERKINESS
`
`HELD REPEAT'
`'
`SUB-SAMPLING
`FREEZE
`
`r
`
`lIMPAiRntxts)
`ACTlYE
`SlilL
`4__._———~—A~'#L—————r
`TV CA¥£RA tlxiD
`
`FANMNG GR ZOOMING
`
`,
`Amt} ‘ SF PICFURE {WORMATION
`
`>
`
`t l l lia lgl ili
`
`
`
`Pll'lURl'QUALIlV
`
`Fig. 13.
`
`{memento coding pefiormancs teatures.
`
`degradation for extreme cases, the interframe coder can
`handle most teleconference scenes.
`Fig. 14 shows a photograph of the NETECré/S termi»
`nal. The encoder and decoder are mounted in a single
`standard 19 in hay withca depth of 24 in:
`
`FUTURE TECHNOLOGIES
`As has been stated, the development of interframe
`coding has reached a point where it is finding real appli-
`cations in television transmission. However, the future
`holds still further innovation. LSl technology'lor high-
`speed logic circuits will contribute greatly to reducing the
`size and increasing the economy of hardware.
`Further innovation will be realized in more sophisti-
`cation of
`interlrame coding algorithms by movement
`compensation. Through movement compensation;
`
`Fly.
`
`14. Photograph olftho Intomame encedcrfdecoder,
`NETEC~B/3._,
`
`PMC3683134
`
`July 1980
`
`the quantization step-size control. At 3 Mbits/s; the
`picture is inquantization control most of the time. How-
`ever,
`if the picture includes some large motion: field
`repeating may operate, causing a slight jerkiness. Sub-
`sampling and picture freezing occur when the television ’
`camera is panned or When television signals are switched
`from one camera to theipther, Degleasing the transmis-
`ysion bit rate results in restricting the range of motion
`which can be transmitted. However, even allowing such
`
`Q
`
`“M 63.5y5
`
`.
`
`t
`Z?
`t.
`[A
`
`Mike
`lice.
`3m Eric?
`Fig. 12.
`
`/
`/
`9’
`v
`1/”
`told color 4V signal lermnl.
`
`184 SAMPLFS
`
`PMC Exhibit 2025
`Apple v. PMC
`IPR2016-00755
`Page 6
`
`

`

`TABLE I
`Chlnnel Clplcity of Various Scheme over Existidg North American Trauaminoion Links
`WW
`Digital
`Analog
`
`lntertrame
`Coding
`lptra
`Network
`frame
`
`Quaiity Teleconferencing
`Encoding
`l
`‘
`
`32—45
`
`Bit Rate
`(Mbits/s)
`
`‘
`
`Satellite
`‘(60‘64 Mbits/s)
`Teirestrial
`(6 GHz, 78 Mbits)
`(IlGHz.90Mbit’s,l
`D5-3 (45 Mblts)
`IDS-2 (6 Mbits}
`DS-lC (3 Mbits)
`DS-l (1.5 Mbits)
`
`'Movement compensation encoder.
`
`HGVUPCM H
`
`PMCAPL02442628
`
`more efficient predictions will be possible by tracking
`the object movement. The concept of movement oom-
`pensation is not new, but has become a serious area'of
`study in digital video processing [30]-{33]. Several re—
`ports have indicated that a doubling or more of the data
`compression ratio can be expected through these tech-
`niques. Thisgis very attractive from an applications point
`of'view. The T1 carrier rate (1.7544 Mints/s) will become
`sufficient
`to transmit a video teleconferencing signal.
`This provides not only a great advantage in transmission
`cost reduction but also a good match to digital networks
`widely used now. Also, in broadcast tele'vision program
`transmission, movement compensation will be quite
`effective particularly for such severe pictures as those
`taken by camera panning or zooming, and will contribute
`to the reduction of transmission rate down to 15 through
`20 Mbits/s—wi'th excellent picture quality.
`_
`Efficient modulation techniques are also important for
`transmission bandwidth compression as well as efficient
`source coding. 'A conventional satellite transponder can
`transmit about 60 Mbits/s over a 36 MHZ channel (1.6 .
`bit/Hz). For a terrestrial radio system using eight-phase
`PSK modulation, the spectral efficiency is in the range of
`2.2—2.6 bits/Hz. Extensive studies and developments
`aimed at higher spectral efficiency (lbits/Hz] using
`multilevel QAM, etc. are undermiay [35]. When such
`technologies become available. an RF' bandwidth of
`10 MHz or even narrower will be sufficient to transmit
`television program signals with a slight
`increase in
`transmitter power.
`Thus, future advances in pource coding and digital»
`modulation techniques will further enhance the advan-'
`tage of
`J
`l transmissron systems.
`
`‘
`
`r
`
`THIS TECHNOLOGY IS READY
`‘
`FOR APPLICATION
`it has been shown that interftame coding techniques
`provide an effective means to reduce the transmission bit
`rate of video signals without sacrificing picture quality. In
`Table l. the channel capacity of various coding schemes
`over existing digital transmission links is listed. For digi-
`tal satellite, for example, even a single television channel
`cannot be transmitted by conventional PCM, whereas
`one or two channels of television can be transmitted
`by employing efficient coding techniques. However,
`much greater advantage is obtained by use of interframe '-
`coding. Three network quality television signals can be -
`carried through a transponder.
`For teleconferehcing applications, a single satellite
`transponder can manage up to twenty simultaneous con
`tereffce signals on a TDMA basis, if one uses a 3 Mbit/s
`interlrame coder. The3 Mbit/s video signal is compatible
`with-the North American digital hierarchy. It can be car-
`ried over two Tl lines or. a single Tl—C line for local dis
`tribution and will be compatible with digital
`toll links
`multiplexed up to the 05-3 45 Mbit/s rate and beyond.
`Our satellite experiment has demonstrath excellent
`performance for network digital television application.
`The TRIDEC teleconferencing system has been com
`mercially' used by NTT in Japan since 1979 and the
`NETECb/B interframe encoder is being seriously con-
`sidered for application in North American teleconfcr
`encing applications. During the past decade. interframe
`coding has grown from theoretiraliinfancy to actual
`application level. Together with the growth of video
`transmission demands and nationwide digital transmis-
`sion over satellite, terrestrial microwave. Coaxial rahle
`
`29
`
`IEEE Communications Magazine
`
`PMC3683135
`
`PMC Exhibit 2025
`Apple v. PMC
`IPR2016-00755
`Page 7
`
`

`

`History of Intel-frame Coding Defidapmcnl
`Authots/insrttute/Roefe rencesWM
`Topic
`Kretzmer, BTL [1] _
`Statistics of teievision Signals;
`Taki et GLLUniv. Tokyo [2}
`Frame correlation study
`Seyler. APG [3}
`Frame difference signal statistics
`- Rocco. Pottch Milann [30]
`Study of bandwidth compression utiirzing frame correiatioe and movement
`compensation
`‘1
`Conditional replenishment PCM
`Frame difference coding for i MHZ monochrome picturephrme Signals
`interfrarne coding for 4 MHz coior tetemsro'n Signals
`Adaptive prediction (movement compensation) study
`15 Mbit/s code: of 1 MHz monochrome videotelephone signaes
`TRIDEC development
`L
`NETEC»22H development
`NEFEC-22H satellite transmtsston experiment and ABS MUX deveiopmem
`1.5 Mbit/s interframe Coding of 525 tine. monochrome television
`Movement compensation experiment for broadcast TV signals
`NETECfi/S terminal development
`TRIDEC commercial service
`30 Mbit/s intertield codmg
`Practical algorithm of movement compensation
`
`PMCAPL02442629
`
`Mounts, BTL [43
`Condyler‘ (545‘, BTL t5]
`imuma et ah. NEC [131,115]
`Haskeil. BTL {31]
`Yasuda et at. NTT [23}
`Yasuda et a], NTI‘ [19].[ZG]
`lshiguw at u.’.. NH.) 124}
`KmmhoetmuNECIBHJZQ'
`Hasked at u.’.. BTi. I21]
`Nmomrya. NHK [34f
`Kaneko. of 01, NEC. NW (18]
`
`Hatori er at, KDD
`Netravah et ul, BTL [33]
`
`Prediction error signal E given by
`
`m
`,7
`E=XA§
`is quantized for transmission. The quantized prediction:
`error is added to prediction signal AX to produce locallyv
`decoded signal Y. Prediction signal X is obtained through
`prediction function Ptz) as follows:
`'
`
`1“!
`A
`X= Z ar'Y,
`i=1
`
`.
`
`(3)
`
`where ar’s are prediction coefficients, Y,’s are time
`sequence of locally decoded signal Y’, and z is the
`z-transform operator."
`‘
`h
`.
`l
`A simplest
`form of predictive coding is differential
`PCM (DPCM) which uses only the previous sample in
`the scan line as the prediction, that'is, N = 1 and a. = 1
`in (2) and (3). For composite NTSC color television
`signals, efficient prediction is made by using plural sam~
`pies (for example, N = 4) in the some scan line (higher--
`order DPCM). Two-dimensional prediction uses sam
`plea: in the previous lines as well as those in the sameiine.
`providing slightly higher efficiency than onedimensional
`prediction. lnterframe coding further uses samples in
`the previous frame. Movement compensation i$/,L20li~
`sidered to be adaptive prediction in which samples VHS
`and coefficients aft; in interframo coding are adaptively
`changed depending on the picture movement to mini-
`mize the prediction error.
`7
`,t
`:
`
`PMC3683136
`
`and optical fiber babies, interftgame coding technology
`will no doubt contribute to new digitai television services,
`
`7
`
`/ 2
`
`APPENDIX
`’3
`X
`'PREDICTIVE conmé
`Data compression of television signals is achieved by
`redundancy removal lot
`television signats. A5 is well
`known, a television signathas-strong correlation be-
`tween picture elements. For example, neighboring pic-
`ture elements have nearly equal amplitudes in most
`cases. This leads to the fact that the difference signal has
`smaller amplitude distribution than the signal
`itself.
`Therefore, by transmitting the element difference infor»
`mation instead of the element amptitude.
`the. signal
`power, or the amount of information to be transmitted
`can be greatly reducedThis is a simple exampie of
`redundancy removal.
`'
`i
`I
`A generalized form of such differential technique is
`“predictive coding.” The principle of it is shown in Fig. 15.
`Input signal X is compared with predicted signal
`
`Transmission
`line
`
`1:
`$2» .
`
`m coder
`
`"Q
`
`uantum
`
`Predictor
`
`Enemy
`
`5
`
`Fig. 15.
`
`Pr‘lrfciplo of predictive coding.
`
`July 1980
`
`PMC Exhibit 2025
`Apple v. PMC
`IPR2016-00755
`Page 8
`
`

`

`_\
`
`{27] J. M. Fraser er UL. “Overall clidtaclexstics ole TASl," BellSysl.
`Tech. J., yol41, July 1962
`‘
`(28] ll. Kaneko. T. lshiéurol’ and M. Smyama, “Digital trails‘ttiission
`through the satellite," presented‘fiat
`lNTELCM 77 EXPUSlllUIl.
`Atlanta, GA’, Oct 1977.
`,.,
`[29; T, Koga, Y. Mirna, K. linuma, and T. lshiguro, “Statistichl analysis
`of NETEC-ZZH system performance.” in'ch. Int. Cont. Comr
`rmm fmi 2, Boston MA, June 1979, pp 2371.23.75.
`F Rocra, “Television bandwidth compressmn utiltzn‘tg frame-to-
`trame correlation and movement compensation," in Proc. Slump.
`Rcrure Bandwidth Compressron, Apr 1969. New York. Garden
`and Breach. 1972
`B, G Haskell. "Entropy measurements tor nonadaptive and
`adaptive, trameto frame.
`linear predictive coding of video tele-
`phone signals,” Bell Syst Tech J. vol. 54, iap. 115571174,
`Aug. 1975.
`'
`‘ Y Ninomtya, “Motion correction for interlrame coding systems,"
`Tech. Group image Engsneering (1E). lECE Japan, Paper lE78-6,
`May 1978.
`‘
`‘ A N , Netravaii andJ. D. Robbins. “Motion compensated tel?»
`i
`Mar
`193’9.
`wiston codinngar: l," Bell SystTech. J., vol.
`pp. 6315670§
`‘r’ Halon,
`Murakami, and ii. Yamamoto. “30 Mbits/s codec
`for NTSCACTV by interfield and intrafiszld adaptive predction,” '
`in Fror
`lnt, Con! Comm-uni. Boston. MA, June 1979. pp.
`23.? 1‘5
`
`‘
`
`“Soccial lssue on Digital Radio," IEEE Trans. thmmun, uni
`COM 27, Dec, 1979
`"
`
`Hilashi Kaneko recetved the B.S.‘degree
`:3". electrical engineering in 1956 from the
`University of TUE-(w. Tokyo. Japan, the
`degree in electrical engineering in 1962mm
`:he University of Calilo‘r‘dia, Berkeley, CA,
`and the Dr. Eng. degree from the University of
`Tflkfi} :n 1%7.
`[n 1956 he jomed the Central Research
`Laboratories, Nippon Electric Company/
`where he worked on digital communications.
`particularly satellite communications, etc.
`Meanwhile, from 1960 through 1962 he studied at the Electronics ,
`Research Laboratory, University of California at Berkeley. as a
`Research