`
`
`
`
`Volume 8, No. 5, Same-miner 19%
`
`
`
`PSSN 6923-5965
`
`
`8l6l475—568l19961
`
`
`
`
`
`Sit? i‘\¥\l_ PRUI' ‘I’SSI \‘(r
`
`IMAGE
`
`COMMUNICATION
`
`CONTENTS
`
`
`
`
`
`
`
`
`
`Papers
`
`
`
`J.T. Virtuma and 83’. Vail-i
`
`
`
`
`
`Vector quantization with hierarchical classification of sub-blocks....................
`
`
`
`
`
`
`
`
`G. Keesmain, R. Hellinglhuizen. F. Hoeksema and G. Heideman
`
`
`
`
`
`
`
`
`
`Transcod-ing of M-PEG b-itstreams ......................................
`
`
`
`
`
`
`
`Y.K. Kim and J.B. Ra
`
`
`
`
`
`Multiple shell structured hypercube feature maps for vector quantization of images ........
`
`
`
`
`
`
`
`
`
`
`
`L. Wu. J. Ben-ois—Pinea-u, P. Dezlalg-n-es and D. Barbs
`
`
`
`
`
`
`
`
`
`Spawn-temporal segmentation of image sequences for object-oriented l-ow bit-rate image coding
`
`
`
`
`
`
`
`
`
`
`M.K.Nl Schoyer and PW. Verbal-ck
`
`
`
`
`
`Block position dithering in OCT-coded sequences .............................
`
`
`
`
`
`
`
`P. Fr'anti, T. Ka-ukoranta and O. Nevalainen
`
`
`
`
`
`
`
`0n the design of a hierarchical ETC-V0 compression system ......................
`
`
`
`
`
`
`
`
`Author index in! Volume 3 ............................................
`
`
`
`
`
`
`Cu-mMa-tive contents of Volume 5 .......................................
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`r”
`
`LAST NUMBER OF THIS VOLUME
`
`
`
`
`
`
`
`I’lnswl'x
`
`lughllnn:
`
`A; NDPHL'Llllnua
`
`A publication—of the European Association for Signal Frocéséing (EU-RAélP) _
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Page 1 of 23
`
`GOOGLE EXHIBIT 1005
`
`

`

`SIGNAL PROCESSING
`
`
`
`
`
`
`IMAGE COMMUNICATION
`
`
`
`Theory. Techniques 8: Applications
`
`
`
`
`
`A publication of the European Association for Signal Processing (EUFtASIP)
`
`
`
`
`
`
`
`
`
`Editor-iwChiel
`
`Leonardo CHIARIGLIONE
`
`
`Centre Sludl e Laboraten
`
`
`
`
`Telecommunicazionl (CSELT)
`
`
`
`Via Guglielmo Reuse-Homo“, 214
`
`
`
`
`I401“ Torino, Italy
`
`
`
`
`Telephone: I It) 2285 120
`
`
`
`Teletext (l 1) 2286 299
`
`
`
`
`Editorial Board
`
`
`J. Bremond (Univ. Delft. The Netherlands)
`
`
`
`
`
`G. Boerger (HHI. Germany)
`
`
`
`
`E. Dubois (Univ. Quebec. Canada)
`
`
`
`
`B. Girod (Erlangen. Germany)
`
`
`
`
`
`
`
`
`
`
`
`Guest Editors
`
`
`P. Aigrain (IFIIT, France)
`
`
`
`
`SA Benton (MIT. USA)
`
`
`
`
`G De Haan (Philips. The Netherlands)
`
`
`
`
`
`T Fujll (NTT. Japan)
`
`
`
`
`J. Guichard (CNET, France)
`
`
`
`
`J Hamasakl (Univ. Tokyo. Japan)
`
`
`
`
`
`J Johann (Deutsche Bundespost. Germany)
`
`
`
`
`T. Kunla (NHK. Japan)
`
`
`
`
`D. LeC-iall (C-Cube. USA)
`
`
`
`
`a. Llu (Princeton Univ. USA)
`
`
`
`
`
`G. Momson (BT Labs. United Kingdom)
`
`
`
`
`
`D. Narasimhalu (Singapore)
`
`
`
`Ft. Nicol (BT Laboratories)
`
`
`
`
`3. Okubo (NTI'. Japan)
`
`
`
`
`T. Omachi (NEC. Japan)
`
`
`
`
`A.P. Pentland (MIT. USA)
`
`
`
`M. Shibata (NHK. Japan)
`
`
`
`
`T. Sikora (HHI. Berlin)
`
`
`
`
`W. Verbiest (Belgium)
`
`
`
`E. Vlso‘rlo (Chromatic Research Corporation. USA)
`
`
`
`
`
`
`
`CN Judlce lBeIL USA)
`
`
`
`
`J.-I(. Kim (KAIST. South Korea)
`
`
`
`
`
`AB Lippman (MIT. USA)
`
`
`
`
`
`H -G Musrnann (Univ Hannover, Germany)
`
`
`
`
`
`D. Nasse (CCETT. France)
`
`
`
`
`Y Nlnornlya (NHK. Japan)
`
`
`
`
`D Pearson (Univ. Essex. United Kingdom)
`
`
`
`
`
`
`F Pereira (IST. Portugal)
`
`
`
`
`L. Stenger (Fl 03?. Germany)
`
`
`
`
`
`H Tominaga (Waseda Univ.. Japan)
`
`
`
`
`M, Tanlmotd (Nagoya Untv., Japan)
`
`
`
`
`
`M, Vetterli (Lausanne, Switzerland)
`
`
`
`
`LT. Wu (ITRI. Taiwan)
`
`
`
`
`H Yasuda (NTT. Japan)
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`TV and Advanced W
`
`
`
`Broadcasting
`
`Image Storage and Retrieval
`
`
`
`Graphic Arts
`
`
`Electronic Prtnting
`
`
`
`
`
`
`Imaging Technology
`
`Display Technology
`
`
`VLSI Processors for
`
`
`
`image Communications
`
`
`
`
`
`
`
`
`
`IMAGE COMMUNI-
`Editorial Policy. SIGNAL PROCESSING:
`
`
`
`
`
`
`CATION is an international (oumal tor the development ol the
`
`
`
`
`
`
`
`
`
`
`theory and practice of image communication. Its primary oblec-
`
`
`
`
`
`
`
`
`
`tlves are the Iollowing:
`
`
`
`
`To present a tumor tor the advancement of the theory and
`
`
`
`
`
`
`
`
`
`
`
`practice of image communication.
`
`
`
`
`To stimulate cross-lenilization between areas similar in nature
`
`
`
`
`
`
`
`
`which have traditionally been separated. for example. various
`
`
`
`
`
`
`
`
`aspects ol visual communications and information systems.
`
`
`
`
`
`
`
`— To contribute to a rapid information exchange between the
`
`
`
`
`
`
`
`
`
`Industrial and academic environments.
`
`
`
`
`The editorial policy and the technical content at the journal are the
`
`
`
`
`
`
`
`
`
`
`
`
`responsibility of the Editor—in‘ChieI and the Editorial Board. The
`
`
`
`
`
`
`
`
`
`Journal is sell-supporting from subscription income and contains
`
`
`
`
`
`
`
`
`a minimum ambunt of advertisements. Advertisements are subject
`
`
`
`
`
`
`
`
`to the prior approval of the Editor-in-Chiet. The (ournal welcomes
`
`
`
`
`
`
`
`
`
`
`contributions from every country in the world.
`
`
`
`
`
`
`
`
`IMAGE COMMUNICATION
`Scope. SIGNAL PROCESSING:
`
`
`
`
`publishes articles relating to aspects of the design. Implementation
`
`
`
`
`
`
`
`
`and use at image communication systems.
`
`
`
`
`
`
`SIGNAL PROCESSING;
`IMAGE COMMUNICATION Ieatures
`
`
`
`
`
`original research work. tutorial and review articles. and accounts ol
`
`
`
`
`
`
`
`
`
`
`practical developments.
`
`
`
`
`
`
`Membership and Subscription Information.
`
`
`
`
`SIGNAL PROCESSING: IMAGE COMMUNICATION (ISBN 0923
`
`
`
`
`
`5965) is published in one volume (six issues) a year. For 1996
`
`
`
`
`
`
`
`
`
`
`
`Volume 3 is scheduled for publication. Subscription prices are
`
`
`
`
`
`
`
`
`available upon request from the publisher, Subscriptions are ac-
`
`
`
`
`
`
`
`
`cepted on a prepaid basis only and are entered on a calendar year
`
`
`
`
`
`
`
`
`
`
`
`basis, Issues are sent by surface mail except to the following
`
`
`
`
`
`
`
`
`
`
`
`countries. where air delivery (S.A.L.
`. Sudace Air Lilted) is en-
`
`
`
`
`
`
`
`
`
`
`sured: Argentina. Australia. Brazn. Canada. Hong Kong. India.
`
`
`
`
`
`
`
`
`Israel. Japan. Malaysia. Mexico, New Zealand. Pakistan. People's
`
`
`
`
`
`
`
`
`Republic at China. Singapore. South Africa. 50th Korea. Taiwan.
`
`
`
`
`
`
`
`
`
`Thailand. USA. For the rest of the world. airmail charges are
`
`
`
`
`
`
`
`
`
`
`
`available upon request. Claims tor missing issues will be honoured
`
`
`
`
`
`
`
`
`
`
`lree oi charge it made wrthin six months afterthe publication date
`
`
`
`
`
`
`
`
`
`
`
`of the issues. Mail orders and inquiries to: Elsevier Science B.V..
`
`
`
`
`
`
`
`
`
`
`
`Journals Department. P.O. Box 21!. 1000 AE Amsterdam. The
`
`
`
`
`
`
`
`
`
`
`Netherlands. For lull membership inlormation ol the Association.
`
`
`
`
`
`
`
`
`possibly combined with a subscription at a reduced rate. please
`
`
`
`
`
`
`
`
`
`
`Subjects. Subject areas covered by the (comet include:
`
`
`
`
`
`
`
`
`contact: EURASIP. PO. Box 134. CH-IOOO Lausanne 13.
`
`
`
`
`
`
`
`
`
`TV. HDTV and (SD-TV systems
`Image Transmission
`
`
`
`
`
`
`
`Switzerland.
`Visual Science
`Interactive Image
`
`
`
`
`
`Communication
`Image Coding
`
`
`
`
`.
`1996 Elsevier Science B.V. All rights reserved
`
`
`
`
`
`
`
`No part of this publication may be reproduced. stored in a retrieval system or transmitted In any lorrn or by any means, electronic. mechamcol. photocopying, recording or
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`othenmse. without the prior permission of the publisher. Elsevier Scrence B.V.. Copyright and Permissions Department. PO Box 521, 1000AM Amsterdam. The Netherlands
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Special regulations (or authors Upon acceptance at an article by the (denial. the euthotls) will be asked to transler copyright ol the ample to the publlsher. This transfer wttl
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`ensure the modal posstble dissemination o! intermalion.
`
`
`
`
`
`
`
`Special regulations for readers In the USA. This )oumat has been registered With the Copyright Clearance Center. Inc Consent Is given tor copying of articles for personal or
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Internet use. or lot the personal use of specific clients This mneent is given on the common that the copier pays through the Center the percopy tee stated in the code on the
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Iirst page of each article lorcopying beyond that ponmtled by Sections It)? or we at the US Copyright Law The appropriate fee should be lorwardedwrth a copy of the lirst page
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`ol the edicts to the Copyright Clearance Center. Inc . 222 Rosewood Drlva. Danvers. MA 09 t23. USA. It no code appears In an article. the author has not given broad consent to
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`copy and pennissron to copy must be obtamed directly from the author The lees Indicated on the first page of an amok; in this issue will apply retroactively to all articles
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`published In the (oumal. regardless ol the year of publication. This consent does not extend to other lands at copying. such as lor general distnbutlon. resale. advertismg and
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`promotion purposes. or tor creating new collective works. Special written permission must be coloured from the publisher tor such copying.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`No responsibility is assumed by the Publisher tor any rnrury andror dampers) persons or property as a manor ol products liability. negligence or otherwise. or Iron-l any use or
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`operation of any methods. products. Instructions or ideas contained In the material herein.
`
`
`
`
`
`
`
`
`
`
`
`
`
`Although all advertising material Is elpected to oonlonn to ethical standards. inclusion In this publication does not constitute a guarantee or endorsement of the qualrty or value
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`oi such product or ol the claims made at it by its manufacturer.
`
`
`
`
`
`
`
`
`
`
`
`
`as» The paper used in this publication meets the requirements of ANSI.=NISO 239.48-1992 (Permanence of Paper).
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Published 6 times a year
`
`
`
`
`
`
`09235965196615.00
`
`
`
`
`Printed in The Netherlands
`
`
`
`
`
`
`Page 2 of 23
`
`Page 2 of 23
`
`

`

`
`
`
`» .
`‘-‘ -": arr-r:
`
`'{Cléi’
`
`.—
`
`
`
`
`Slgnal Processing: IMAGE COMMUNICATION
`
`
`
`
`
`
`Volume 8, No. 6, September 1996
`
`
`
`
`
`‘
`
`
`
`CONTENTS
`
`
`
`
`Papers
`
`'4 4
`
`._ .__,t._.1
`" sum i
`
`
`
`
`
`"C'vFMl
`‘
`
`
`
`
`
`
`wwmv \hnn (tn: '1!
`
`
`.,4
`
`
`
`
`A 'JG 2995 :7
`
`
`
`
`J.T. Virtamo and S.T. Valli
`
`
`
`
`
`
`
`
`
`
`
`
`Vector quantization with hierarchical classification of sub-blocks .............
`
`
`
`
`
`
`
`
`
`G. Keesman, R. Hellinghuizen, F. Hoeksema and G. Heideman
`
`
`
`
`Transcoding of MPEG bitstreams ................................
`
`
`
`Y.K. Kim and J3. Ra
`
`
`
`
`
`
`
`
`
`
`
`
`Multiple shell structured hypercube feature maps for vector quantization of images.
`
`
`
`
`
`
`
`
`
`L. Wu, J. Benois-Pineau, P. Delagnes and D. Barba
`
`
`
`
`
`
`
`
`Spatio-temporal segmentation of image sequences for object-oriented low bit—rate
`
`
`image coding ............................................
`
`
`
`
`
`M.K.N. Schoyer and P.W. Verbeek
`
`
`
`
`
`
`Block position dithering in DCT-coded sequences ......................
`
`
`
`
`
`
`
`P. Franti, T. Kaukoranta and O. Nevalainen
`
`
`
`
`
`
`
`
`On the design of a hierarchical BTC—VQ compression system ...............
`
`
`
`
`
`Author index of Volume 8 .....................................
`
`
`
`
`
`
`Cumulative contents of Volume 8 .................................
`
`
`
`
`
`
`
`475
`
`
`481
`
`
`501
`
`
`513
`
`
`545
`
`
`551
`
`
`563
`
`565
`
`
`Advance tables of contents can be retrieved at
`
`
`
`
`
`
`
`
`
`
`http://www.elsevier.nl/iocate/eee.
`
`
`
`Page 3 of 23
`
`Page 3 of 23
`
`

`

`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`1MAGE
`
`
`(.H‘vl,\1l\l( \|l()\
`
`
`
`
`'I'I‘Lmscudmg 01‘
`
`
`
`
`.\-I PM} hilslrcmns
`
`
`
`
`(Ecrljgm chumn
`
`
`
`
`Rulwrl
`
`
`
`
`
`
`
`llcllinglmi/L‘ni'. l-uMc llnclwunLUL (:cci'l Iimiunmn"
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`\ hx‘trzn’i
`
`
`
`
`
`
`
`"
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`L
`
`
`
`
`
`
`i
`l‘,1»“\:
`.
`
`.
`
`
`ilih Ulil‘cl «ham‘nw- 'JJ.‘ gm-Hcm w! Hanu'uxil‘
`
`
`
`
`
`
`.1 muslin t:‘.L:LnI11:I~.L51 '\.[]’l (J -.-~|'1;'»;-qw.mi
`'.\-.|-.'n-\
`
`
`
`
`
`
`
`HAUL-1:1wig; UL
`«A: llmH.wi..
`
`
`
`
`
`
`‘
`Il"~“’
`M,
`
`
`
`
`
`
`1‘.
`‘ui-
`
`
`
`:
`
`
`
`
`\
`
`‘
`
`
`
`
`
`.
`
`.
`
`
`
`
`
`
`
`.
`
`
`
`K
`
`
`
`
`
`‘
`
`
`
`
`
`
`
`
`
`
`
`1
`
`
`
`
`
`
`
`
`~
`
`-
`
`
`
`r
`
`
`
`
`
`
`
`
`
`
`
`1'
`
`~
`
`
`
`
`‘
`
`
`
`
`
`|
`
`‘
`
`
`
`
`u
`, ”_
`
`
`
`~
`
`
`
`i
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`\.
`
`
`.
`
`‘
`
`
`
`~
`
`
`
`
`
`
`
`
`
`
`:
`
`\
`
`
`
`‘
`
`x
`
`
`
`‘.‘
`
`
`
`\
`
`
`
`'
`
`
`
`
`
`
`"
`
`
`‘
`
`
`
`
`
`
`
`‘I
`
`.
`
`\_
`
`
`
`
`‘- u
`
`
`.
`
`
`
`
`
`
`
`
`
`
`
`
`'
`
`
`
`
`
`
`
`
`
`\
`
`
`
`
`
`‘
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`I
`
`
`
`
`
`
`
`\
`
`‘
`
`~\
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
` H ‘
`
`...
`
`
`
`
`.
`
`‘,..
`
`,_\
`
`‘.:..._.;~
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`mm :\ \,.1Hn: .1 "mm. will;
`Hmm‘mlm:
`H
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`'1‘. 'th pun-1 Ihc VPCL-H'k
`51; mm cum-shivm xx
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`(SP0: 'S'n LI SIM.) MEI NEH/M103 Aq Penewd Sq Aw lelJSJEU-l 5N1
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`‘
`
`
`
`
`
`
`a
`
`
`‘
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`..
`
`
`.
`
`
`
`
`
`
`
`
`
`
`‘
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`.
`
`
`
`
`
`
`
`’
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`‘
`
`
`
`
`'
`
`
`
`
`
`\
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Page 4 of 23
`
`

`

`(1 Km .rvmm or .u’
`
`
`
`
`
`Swim! Fri-resting himur- ('nnmitmit‘tmrm N il‘l'm; J.\I
`
`
`
`
`
`
`
`
`
`firm
`
`
`4R:
`
`ErEncoder
`
`
`D=Decooer
`
`
`
`T=Transcoder
`
`
`
`
`Videotn l "_ R1
`,
`.
`n
`
`‘ _ R2
`4.4
`
`‘
`
`
`
`i,
`
`
`
`‘
`
`theoont
`
`
`
`
`
`
`
`Fig.
`
`
`l. Basic configuration of a system including a transcoder.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`discussed. Since MPEG video compression [see eg.
`
`
`
`
`
`
`
`
`
`
`{2. 3. 5. 6. 8. 10. 12]) is the current mainstream in
`
`
`
`
`
`compression technology this paper will concentrate
`
`
`
`
`
`
`
`
`
`
`on the transcoding of MPEG signals into MPEG
`
`signals.
`
`
`
`
`
`
`Basically. the situation with transcoding is as
`
`
`
`
`
`
`
`
`shown in Fig. 1: an encoder compresses the incom-
`
`
`
`
`
`
`
`
`
`
`ing video signal at a hit-rate of R, (Mbitv‘sj, then
`
`
`
`
`
`
`
`
`this compressed signal
`is converted into a com-
`
`
`
`
`
`
`
`pressed format of a lower bit-rate R3 [Mbitgsl and
`
`
`
`
`
`
`
`finally a decoder decompresses the incoming signal
`
`
`
`
`
`
`and displays the resulting video signal. Transcod-
`
`
`
`
`
`
`
`
`
`ing may occur in situations where one means of
`
`
`
`
`
`
`
`signal transport interfaces another means of signal
`
`transport.
`
`
`
`
`
`
`Essentially. a transcoder consists of a cascaded
`
`
`
`
`
`
`
`
`
`
`decoder and encoder. as shown in the lower part of
`
`
`
`
`
`
`
`
`
`
`Fig. I. In the situation described in this paper the
`
`
`
`
`
`
`
`encoder embedded in the transcoder is only com-
`
`
`
`
`
`
`
`pressing at a lower bit-rate than the incoming bit—
`
`
`
`
`
`
`
`rate. so there is no other reformatting involved.
`
`
`
`such as resampling.
`
`
`
`
`
`
`
`[n the literature the problem of transcoding
`
`
`
`
`
`
`
`
`seems fairly uncovered.
`the problem of
`in [9]
`
`
`
`
`PCM-ADPCM-PCM lossless tranSCoding is dis-
`
`
`
`
`
`
`
`
`cussed. Morrison et al. [7] describe a transcodcr
`
`
`
`
`
`
`
`
`that has a limited complexity for video signals
`
`
`
`
`
`
`
`Compressed with hybrid DCT systems. Morrison et
`
`
`
`
`
`
`
`
`
`al. compare the picture quality in a situation when
`
`
`
`
`
`
`
`transcoding from bit-rate R,
`to R; with the situ-
`
`
`
`
`
`
`
`
`
`
`ation where the video signal is directly compressed
`
`
`
`
`
`
`
`at R2. Their measurements indicate an extra loss
`
`
`
`
`Page 5 of 23
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`I dB for H.261
`
`
`
`compressed
`
`
`
`
`
`
`
`
`
`
`
`of approximately.
`
`signals.
`
`
`
`
`
`
`Transcoding by means of cascading a decoder
`
`
`
`
`
`
`
`
`
`and an encoder may lead to several types of prob-
`
`
`
`
`
`
`
`
`
`lems. In this paper two problems are described. The
`
`
`
`
`
`
`
`
`first problem is that of complexity. It will be shown.
`
`
`
`
`
`
`
`
`
`
`
`
`in a way similar to that of Morrison [4]. that the
`
`
`
`
`
`
`complexity can be reduced significantly. The trans-
`
`
`
`
`
`
`
`coder discussed in this paper is. however. slightly
`
`
`
`
`
`
`
`less complex than that of Morrison. Another differ—
`
`
`
`
`
`
`
`
`
`
`ence with respect
`to Morrison et al.
`is that the
`
`
`
`
`
`
`
`
`
`transcoder discussed in this paper will be applied to
`
`
`
`
`
`MPEG-compressed signals instead of H.26l-com-
`
`
`
`
`
`
`
`
`pressed signals as in the case of Morrison.
`
`
`
`
`
`
`
`
`The second problem is that of performance, that
`
`
`
`
`
`
`
`is, picture quality at a given bit-rate. The introduc— '
`
`
`
`
`
`
`
`
`tion ofa transcoder in the transmission chain may
`
`
`
`
`
`
`
`
`introduce extra losses as compared to compressing
`
`
`
`
`
`
`
`
`
`directly. called direct coding. to the lowest bit-rate.
`
`
`
`
`
`
`
`
`
`This can be investigated via two approaches: (1)
`
`
`
`
`
`
`
`theoretical analysis and (2) computer simulation.
`
`
`
`
`
`
`
`Signals passing through transcoders are typically
`
`
`
`
`
`
`
`quantized twice. This cascaded quantization is ;
`
`
`
`
`
`
`
`
`
`
`a source of performance loss in the case of trans-
`
`
`
`
`
`
`coded signals. Besides theoretically analyzing cas-
`
`
`
`
`
`
`
`
`caded quantization it
`is also possible to simulate
`
`
`
`
`
`
`
`
`a complete transmission path including a trans-
`
`
`
`
`
`
`
`
`
`coder. The latter does not involve any model ab:
`
`
`
`
`
`
`
`
`
`straction. but may yield less insight into the mecha-
`
`
`
`
`
`
`
`nisms which actually cause performance loss.
`
`
`
`
`
`
`
`
`
`_
`‘
`This paper is organized as follows. Section 2 will
`
`
`
`
`
`
`
`
`discuss the problem of reducing the complexity of If
`._
`
`
`
`
`
`
`
`
`
`the transcoder. The problem of picture quality loss
`
`
`
`
`
`
`
`
`
`
`will be discussed in the following three sections.
`
`
`
`
`
`
`
`
`
`
`A typical source of performance loss in a system
`
`
`
`
`
`
`
`
`including transcoding is the double occurrence of
`
`
`
`
`
`
`
`
`
`a quantization operation. The first step is to ana-
`
`
`
`
`
`lyze
`cascaded
`quantization
`straightforwardly,
`
`which will be discussed in Section 3. This section
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`will present an analysis for single coefficients so
`
`
`
`
`
`
`
`
`
`that insight into the basic mechanisms can be ob-- '
`
`
`
`
`
`
`
`
`
`
`tamed and this section will present an analysis for,
`the combination of 64 coefficients. Section 3.4 will
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`discuss the second step, which is to enhance the.
`
`
`
`
`
`
`quantization model with prediction loops. Results!"
`
`
`
`
`
`
`obtained in computer simulations of a system in-g
`
`
`
`
`
`
`
`
`
`eluding transcoding will be discussed in Section 4.
`
`
`
`
`
`
`
`
`
`7 Finally. the paper will be concluded in Section 5.
`
`'
`
`
`
`
`
`'V
`
`Page 5 of 23
`
`

`

`(if Kwsinun r! u!
`
`
`
`
`
`
`
`.‘t'tgmt'i' f't‘nu'ssntt:
`
`
`
`
`liturgy t
`
`
`'umiirmm‘ulmn 45' IIUWI.‘ 4.” firm
`
`
`
`
`
`
`.JHi
`
`
`
`
`
`2. Complexity of transcodcr
`
`
`
`
`
`
`
`
`In the introduction a transcodcr was pictured as
`
`
`
`
`
`
`
`a cascaded decoder and encoder. Straightforward
`
`
`
`
`
`
`
`
`cascading of the decoder and encoder. however.
`
`
`
`
`
`
`
`
`
`may lead to far too complex transcoders. In this
`section it will be shown that the transcoder can be
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`much simpler. i.e. the number oftllDCTs and the
`
`
`
`
`
`
`
`
`
`amount of memory required can be reduced by
`
`
`
`
`
`
`
`
`combining the decoder and encoder in a more
`
`
`
`
`
`
`
`
`cfiicient way. The approach will be to cascade a de-
`
`
`
`
`
`
`
`
`coder and an encoder and then analyze which ele-
`ments of the two can be combined.
`
`
`
`
`
`
`
`
`
`
`
`
`The transmission chain being studied comprises
`
`
`
`
`
`two cascaded encoding and decoding operations.
`
`
`
`
`
`
`
`The variables associated with the first encoding and
`
`
`
`
`
`
`decoding operation will have superscript (It and
`
`
`
`
`
`
`the variable associated with the second encoding
`
`
`
`
`
`
`
`and decoding operation will have superscript {2). ln
`
`
`
`
`
`
`
`
`
`system blocks subscripts l and 2 will be used to
`denote this difference.
`
`
`
`
`
`
`
`
`
`
`
`
`is
`Fig.2 shows the outline of a decoder that
`
`
`
`
`
`
`
`
`present
`in the transcoder. First
`the bit-stream is
`
`
`
`
`
`
`
`
`decoded (VLD). yielding the value of the quantized
`
`
`
`
`
`
`coefficients. next
`these quantized coefficients are
`
`
`
`
`
`
`
`dis-quantized tDQfi and put through an inverse
`
`
`
`
`
`
`
`discrete cosine transform tIDCTt. This process
`
`
`
`
`
`
`
`
`
`yields the residual signal RLl’ofthc picture IL“. The
`
`
`
`
`
`
`
`
`picture can be reconstructed by adding the predic-
`
`
`
`
`
`
`
`
`tion which results from the previously decoded
`
`
`
`
`
`
`
`picture IL'J, by applying motion compensation
`
`(shift).
`
`
`
`
`
`
`
`
`
`
`
`
`Let V}, denote the vector field of picture it. in fact.
`
`
`
`
`
`
`each component ofthe vector represents a displace-
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`ntcnt fora pixel. lithe position ofa pixel in a pic-
`
`
`
`
`
`
`
`
`
`turc is denoted by x then the displacement is here
`
`
`
`
`
`
`
`
`denoted l',,txi. The inotion—compcnsated signal
`is
`
`
`
`
`
`
`
`
`denoted S[1,‘,'..’[. 1;]
`in Fig. 2 and the operation
`
`
`
`
`
`
`
`
`
`
`S[-] will be referred to as the shift operation. The
`
`
`
`
`
`
`
`motion compensation yields the prediction Pn :
`
`
`
`S[I,‘.“,. R] as follows:
`
`
`
`P,.tx) = S[I,‘."1.It,',]txl
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`According to the MPEG syntax. the displace-
`
`
`
`
`
`
`
`
`
`merits 1“,,tfoor each pixel are equal for pixels within
`
`a macroblock. The macro-block vectors are trans-
`
`
`
`
`
`
`
`
`
`
`
`
`
`mitted in the bit-stream. So. besides quantized
`DCT coefficients. motion vectors are also decoded
`
`
`
`
`
`
`
`
`
`
`
`
`
`from the incoming bit-stream. Furthermore.
`the
`so-cailed macro-block modes are also decoded
`
`
`
`
`
`
`
`
`
`
`
`
`from the incoming bit-streams. These macro—block
`modes contain information about. for instance. the
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`intras‘inter decision or the forwa rd s’back wardj'inter-
`
`
`
`
`polated decision [3. 10].
`
`
`
`
`
`
`
`A significant
`reduction in complexity can be
`achieved if the cascaded encoder uses the same
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`picture types as the incoming signal. So. when a de-
`
`
`
`
`
`
`
`
`
`coded l picture is again coded as an I picture.
`
`
`
`
`
`
`
`
`
`
`a decoded P picture is again coded as a P picture
`
`
`
`
`
`
`
`
`
`
`and a decoded B picture is again coded as a B pic-
`
`
`
`
`
`
`
`ture. the transcoder complexity can be reduced due
`to two factors.
`
`
`
`
`
`
`
`
`
`
`
`
`First ofail. the decoded pictures are not yet in the
`
`
`
`
`
`
`
`
`proper order. That is. although B pictures are de-
`
`
`
`
`
`
`
`coded after the corresponding anchor frames. they
`
`
`
`
`
`
`
`
`are displayed between the two anchor frames. This
`
`
`
`
`
`
`
`
`Iit.)
`n
`!.
`
`
`
`t
`
`
`,
`
`
`video
`
`
`
`
`
`
`
`Hm
`.
`.
`
`7.
`bits
`
`
`
`fl ——~
`VLD ‘_.
`no1
`
`
`i
`
`——a
`
`f IDCT 1;! ~____
`
`+
`l
`
`
`
`Pt|t_s|tli y]__//'
`
`
`
`-1-
`
`
`
`
`
`
`n
`
`
`
`
`n
`
`
`
`Shin — MEM
`
`
`—
`
`A \y"
`
`r
`
`
`
`(It
`tn-l
`
`
`
`
`Fig. 2. Outline of the decoder in the Iranscoder.
`
`
`
`
`
`
`
`
`
`
`
`
`V—w-_._,
`
`Page 6 of 23
`
`Page 6 of 23
`
`

`

`484
`
`
`(i Ki‘iummt v! a!
`
`
`
`
`
`Sign”! f’rmrrruru' image Currtrrrtmit-tmrm N HWM MI 50”
`
`
`
`
`
`
`
`
`
`I
`
`III
`n
`l
`+
`t
`
`~‘ -—I-l
`Video
`_
`
`
`ill
`Fl
`,n
`
`.f
`it
`
`Prediction /'
`
`
`
`v
`Pm 7 3:1“,
`
`loaf-n] -
`n
`*
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`flAyn (motion vectors)
`
`
`
`
`
`Fig. 3. Outline of the encoder in the transcoder.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`so-called picture reordering is performed inversely
`
`
`
`
`
`
`
`
`in the decoder. The picture reordering in the de-
`
`
`
`
`
`
`
`
`coder and the picture ordering in the encoder can-
`
`
`
`
`
`
`
`
`
`
`cel out when the two are cascaded. Here a large
`
`
`
`
`
`
`amount of memory can be saved.
`
`
`
`
`
`
`
`Secondly. most of the decoded information. such
`
`
`
`
`
`
`
`
`as motion vectors and macro~block modes. can be
`
`
`
`
`
`
`
`reused in the cascaded encoder. Especially. the ob-
`
`
`
`
`
`
`
`viated need to estimate motion vectors again red-
`
`
`
`
`
`
`uces the computational complexity of the trans»
`
`
`coder significantly.
`
`
`
`
`
`
`
`
`
`
`At this point there is no need to have decoded
`
`
`
`
`
`
`
`
`pictures available in the transcoder because most of
`
`
`
`
`
`
`
`
`
`the incoming data can be copied from the decoder
`
`
`
`
`
`
`
`
`
`
`to the encoder. In the following it will be shown
`
`
`
`
`
`
`
`
`
`that. because there is no need for decoded pictures.
`
`
`
`
`
`
`
`
`the amount of memory for storing the previously
`
`
`
`
`
`
`decoded pictures can also be reduced.
`
`
`
`
`
`
`
`
`The typical MPEG-2 picture level features tsee
`
`
`
`
`
`
`
`
`
`e.g. [3]] can also be copied from the incoming
`bit-stream to the encoder. This holds for features
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`'pii'turefistrttcture‘.
`'topifield—firsr'.
`such as
`
`
`
`
`peat—firsrsfiefd' and progressive. rume'.
`
`
`
`
`
`
`
`
`
`Fig.3 shows the basic outline of the encoder.
`
`
`
`
`
`
`
`
`the prediction Pl“ is subtracted from the
`First
`
`
`
`
`
`
`
`
`incoming pictures ll". which yields the residual
`
`
`
`
`
`
`
`signal R1,“. This residual signal is put through a dis-
`
`
`‘rv-
`
`Page 7 of 23
`
`l4
`
`
`
`
`
`
`
`
`
`crete cosine transform (DCTl and a quantizer (Q2).
`
`
`
`
`
`
`
`The quantized DCT coefficients are compressed
`
`
`
`
`
`
`
`and then outputted. Also. the quantized DCT coef-
`
`
`
`
`
`
`
`
`ficient are again dequantized and are further fed
`
`
`
`
`
`
`
`
`through an lDCT to yield the reconstructed resid-
`
`
`
`
`
`
`
`
`ual picture Rf,” + EL“. where El,“ denotes the
`
`
`
`
`
`
`
`
`quantization error in the second encoder. i.e. the
`
`encoder in the transcoder.
`
`
`
`
`
`
`
`
`
`
`
`
`
`In the next step the picture IL“ is reconstructed
`
`
`
`
`
`
`
`
`by adding the same prediction as was subtracted
`
`
`
`
`
`
`
`
`earlier in the process. This prediction is obtained by
`
`
`
`
`
`
`
`
`
`applying the shift operation S[-] to the previously
`
`
`decoded picture.
`
`
`
`
`
`
`
`The basic transcoder consisting of a decoder (see
`
`
`
`
`
`
`
`
`
`
`Fig. 2) and a cascaded encoder {see Fig. 3} is shown '_
`
`
`
`
`
`
`
`
`
`
`in Fig. 4. Observe that only the prediction loop of
`
`
`
`
`
`
`
`
`
`the decoder is shown. The transcoder of Fig. 4 can
`
`
`
`
`
`
`
`
`
`
`be simplified by removing one of the two picture
`
`
`stores (Ml-3M}.
`
`
`
`
`
`
`
`
`
`
`
`This can be done using the fact that the shift
`
`
`
`
`
`operator is linear. that is.
`
`‘
`
`_
`
`
`
`$0133, + 5:15.13 =
`
`
`
`
`
`Stir: .. K] + star: .. a]
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`(ll .
`
`
`
`
`
`
`
`
`holds. Using this linearity preperty the prediction.
`
`
`
`
`
`
`
`
`
`in the second encoder. i.e. the encoder within the '
`
`Page 7 of 23
`
`

`

`G. Keesmun e! u[,
`
`
`
`
`
`Signal Processing: Image Community/inn 8 1/996) 48/ 50!)
`
`
`
`
`
`
`
`
`
`
`485
`
`
`
`Point Y
`
`
`
`
`Form X
`
`
`
`
`(“\‘Zn(2)
`:hJHDcrHoz l___—.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`n
`
`=
`
`n
`'(2)
`
`n
`l(1)
`
`/
`o
`n
`(
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
` Slr‘i’vvnF E'léi’r 8!] SElEli’t
`
`EWY
`
`Fig. 4. Cascaded decoder and encoder as transcoder.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`All that is needed to be able to produce the out-
`
`
`
`
`
`
`going bit-stream is the residual RLZ’. According to
`
`
`
`
`
`
`
`
`Eq. (4), this residual Rf,” can be obtained directly
`
`
`
`
`
`by subtracting the motion-compensated quantiz-
`
`
`
`
`
`
`
`ation error E13 1. Hence, only £5,“ need be stored.
`
`
`
`
`
`
`
`
`
`
`
`and the store used for It.” can be dropped.
`
`
`
`
`
`
`
`
`
`Fig.5 shows the block diagram resulting after
`
`
`
`
`
`
`
`
`
`the removal of a picture store. To obtain the coding
`
`
`
`
`
`
`
`
`
`error El,” the input and the output of the coding
`
`
`
`
`
`
`
`path consisting of DCT, quantizer, de-quantizer
`
`
`
`
`
`
`
`and [DCT are subtracted. Consecutively the coding
`
`
`
`
`
`
`
`
`
`error E,” is stored in a memory. The residual signal
`
`
`
`
`
`
`R"j is obtained by subtracting the motion-compen-
`
`
`
`
`
`
`
`sated error S[E'2_’ ,, ,,] from the decoded residual
`
`
`
`RU)
`
`
`
`
`
`
`
`Another simplification can be realized in the
`
`
`
`
`
`
`
`
`
`transcoder of Fig.5 by moving the DCTs and
`
`
`
`
`
`
`
`
`IDCTs around in the block diagram. The DCTs
`
`
`
`
`
`
`
`
`
`and lDCTs can be moved around by using the
`
`
`
`
`
`
`
`
`
`
`linearity property of the DCT. Fig. 6 shows some of
`
`
`
`
`
`
`
`
`
`the possibilities. The top section of the figure shows
`
`
`
`
`
`
`
`
`
`
`
`that in a data fork the DCT (or IDCT) can be
`
`
`
`
`
`
`
`
`
`moved from the main path to the two separating
`
`
`
`
`
`
`
`
`
`transcoder, can be split into two components. One
`
`
`
`
`
`
`
`
`of the two components in the encoder prediction
`
`
`
`
`
`
`
`results from the motion compensation applied to
`
`
`
`
`
`
`the previously decoded picture 1:331. Observe that
`
`
`
`
`
`
`
`
`
`
`
`
`this prediction is first added in the decoder at 'point
`
`
`
`
`
`
`
`
`
`
`
`X‘ in the figure and is then subtracted in the second
`
`
`
`
`
`
`
`
`
`encoder at 'point Y’ in the figure. These two opera-
`
`
`
`
`
`
`
`
`
`tions of course cancel one another. It is only neces-
`
`
`
`
`
`sary to subtract the motion-compensated quantiz-
`
`
`
`
`
`
`
`
`
`ation error. as will be shown in the following.
`
`
`
`
`
`
`
`Using Eq. (I),
`the residual RLZ’
`in the second
`
`encoder can be written as
`
`
`
`
`
`
`
`
`
`
`
`
`
`R1.” = If." -5[15.'—'r + E}. l. V..]
`
`
`
`
`
`
`
`
`= 1".”~SU.‘.'-’1J’5J -S[E.‘.2—’r. Kl
`
`
`
`
`
`
`
`
`
`The decoded picture 15,“ can be written as
`
`
`
`
`1'“: R‘“+S[I}.‘_', V,,].
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Substitution of Eq. (3) into Eq. (2) yields
`
`
`
`
`R”): R“’— S[E§,’.’x, K].
`
`
`
`
`
`
`
`(2)
`
`
`(3)
`
`
`(4)
`
`Page 8 of 23
`
`Page 8 of 23
`
`

`

`4““
`
`
`(5. Kremmm or (1!,
`
`
`
`
`
`Sign”! f’t'ut'tu'xmg.‘ [Hinge ('nntmumumml r‘\' erm 43H 5””
`
`
`
`
`
`
`
`
`
`
`HI
`n
`
`(”’I
`n
`
`from VLD
`
`
`
`
`
`
`I"
`
`+ A
`
`i
`7
`
`‘
`
`
`
`
`
`
`
`
`
`
`
`
`ii;IDCT lat”) —————-- oer—:+Q;
`
`
`
`
`
`
`
`slew v 7
`
`
`
`
`
`
`.n-I ‘n-“4
`
`toVLC
`
`
`
`
`
`
`
`
`ll
`
`
`
`
`
`
`
`Fig. 5. Transcodcr with a reduced amount of memories.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`.—>
`(llDCT
`
`”l
`
`‘—>[ moor —+
`‘ -'
`
`,7 moor
`
`
`
`il—._ (DUCT .
`
`
`
`
`
`i
`
`6% E tapas*4
`L“
`fl
`
`
`
`flfi
`
`(59
`
`
`Ill
`
`t
`sT
`
`--
`
`
`Fig. 6. Moving DCT and lDCT operations around.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`part of Fig. 6 in combination with the property lot
`
`
`an addition.
`_
`
`
`
`
`
`
`
`
`
`Moving the DCTs and [DCTs of Fig 5 around
`
`
`
`
`
`
`
`
`
`
`in this way yields the transcoder of Fig. 7. Obser . .
`
`
`
`
`
`
`
`
`
`
`that the number of DCTs and IDCTs has been:
`
`
`
`
`
`
`
`
`
`reduced from three to two. At this point the 0011i;
`
`.
`
`
`
`
`
`
`
`
`
`paths; the middle section shows that at an addition
`
`
`
`
`
`
`
`
`
`or a subtraction the DCT (or IDCT) olthe compo-
`
`
`
`
`
`
`
`
`
`
`
`nents of the addition can be moved to the DCT (or
`
`
`
`
`
`
`
`
`
`lDCT) of the result. Finally. applying an IDCT to
`
`
`
`
`
`
`
`
`
`a signal that has passed through a DCT equals
`
`
`
`
`
`
`
`
`
`
`a unity transfer. This property is used in the lower
`
`
`
`Page 9 of 23
`
`
`
`
`
`
`Page 9 of 23
`
`

`

`G. Kcexnum v! (1/
`
`
`
`
`
`
`" Signal Processing: Image Communication 8 ([990) 4.?! 450”
`
`
`
`
`
`
`
`
`
`487
`
`
`
`
`to VLC
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`Fig. 7. Transcoding system with a reduced number of DCl's.
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`The whole procedure followed for deriving the
`
`
`
`
`
`
`
`
`
`
`
`transcoder of Fig. 7 is implicitly based on the use of
`
`