`
`US 20060146138Al
`
`(19) United States
`
`(12; Patent Application Publication (10) Pub. No.: US 2006/0146138 A1
`
`Xin et al. Jul. 6, 2006 (43} Pub. Date:
`
`
`(54) METHOD AND SYSTEM FOR
`SYNTHESIZING MULTIVIEW VIDEOS
`
`Publication Classification
`
`(76)
`
`Inventors:
`
`.Iun Xin, Quincy, MA (US); I‘Imin
`Martinian. Waltlunn. MA (US):
`Alexander Behrcns‘ Bueckeburg (DE):
`Anthony Vctro. Arlington, MA (US);
`Iluifang Sun. Billerica. MA (US)
`
`Correspondence Address:
`Patent Department
`Mitsubishi Electric Research Laboratories, Inc.
`20] Broadway
`Cambridge, MA 02139 (US)
`
`(21) App]. No.2
`
`l'UZ92,lfi7
`
`(22)
`
`1.11“}:
`
`No“ 3|}, 2005
`
`Related U,S_ Application Data
`
`(63)
`
`(.‘onlintuttion-in-part of application No. [”015390.
`filed on Dec. 17, 2004.
`
`(5] )
`
`Int. (7].
`(2006.01)
`HIMN 5/225
`(52) U.S. (Tl.
`........................................................ 348207.99
`
`(57)
`
`ABSTRACT
`
`A system and method syntheslzes multwiew VldeOS. Multi—
`\r'iL‘W videos are acquired of a scene with corresponding
`cameras arranged at a poses such that there is View overlap
`between any pair ofcameras. A synthesized multiyiew yideo
`ls generated from the acquired mu ltmew Videos for a virtual
`camera. A reference picture list
`is maintained for each
`current
`frame of each of the nlulliview videos and the
`synthesiaed video. The reference picture list indexes teln-
`poral reference pictures and spatial reference pictures of the
`acquired multiview videos and the synthesized reference
`pictures of the synlhesized multiview video.
`'l‘hen. each
`current frame of the inultiview videos is predicled according
`to reference pictures indexed by the associaled reference
`picture list during encoding and decoding.
`
`Prediction Modes
`
`.
`
`spatial
`temporal
`View synthesis
`intra
`
`4‘0
`
`402
`
`
`Input Video 3
`
`403
`
`
`MCTF/DCVF
`
`412
`
`
`Input Video 1
`r 40]
`411
`
`
`Input Video 2
` High band
`frames
`Decomposition
`
`
`
`
`Input Video 4
`
`404
`
`
`Side
`Information ( 413
`
`UNIFIED 1003
`
`UNIFIED 1003
`
`
`
`1
`
`“
`
`‘1.
`
`":
`‘ i-
`
`scenc
`
`g
`in
`1
`
`,5
`
`3
`5'
`i I.
`
`5) 4
`‘~
`
`101
`
`102
`
`103
`
`'
`
`04
`
`FIG. I
`
`PRIOR ART
`
`61J0Imalls
`
`Encoded
`video 2
`
`122
`
`Encoded
`video 3
`
`123
`
`Encoded
`video 4
`
`124
`
`.—.
`
`100
`
`Encoded
`
`Encoder 1
`
`video 1
`
`121
`
`F‘
`
`—»
`
`
`
`uoueauqnduoulznnddvmama
`
`900E‘9‘[“I'
`
`IV8£l9|7l019003SI]
`
`
`
`211
`
`200
`
`22]
`
`231
`
`-'
`
`
`
`201
`
`202
`
`203
`
`
`
`204
`
`214
`
`252
`
`234
`
`
`
`
`
`FIG. 2
`
`PRIOR ART
`
`
`
`
`
`uogwaglqnduoulznnddvwaned
`
`900K‘9’[nl'
`61.10ZIaaus
`
`IV8£l9|7l0l9003SI]
`
`
`
`302
`
`304
`
`High band
`samples
`
`
`
`
`
`
`6110£Iaaus900z‘9'lnr"OUBNJQHd"oumnddvmawd
`
`
`
`
`
`FIG. 3
`
`PRIOR ART
`
`Low band
`
`samples
`
`IV8€l9|7l0l9003SI]
`
`
`
`
`
`Prediction Modes
`
`
`
`’
`
` 411
`
`412
`
`uoynouqnduogeauddvwaned
`
`
`
`6|10vwallssum‘9'[fll‘
`IV8€l917l0f9002Sf]
`
`spatial
`temporal
`
`Input Video 1
`F‘
`
`
`
`410
`view synthesis
`
`
`intra
`
`
` 401
`Low band
`frames
`
`MCTF/DCVF
`Highband
`
`Decomposition
`
`Side
`Input Video 4
`Information ( 413
`404
`
`
`
`1”)
`
`Input Video 2
`
`402
`
`Input Video 3
`
`403
`
`400
`
`FIG. 4
`
`
`
`Patent Application Publication
`
`Jul. 6, 2006 Sheet 5 0f 19
`
`US 2006f0l46138 Al
`
`time
`
`502
`
`44/ 4,:
`
`’
`
`Ijfii:
`
` 4"
` space
`
`
`
`Patent Application Publication
`
`Jul. 6, 2006
`
`Sheet 6 of 19
`
`US 200610146138 Al
`
`
`
`
`
`700
`
` 711-712
`730
`
`
`Macroblock-
`
`Adaptive ”I Encoder
`
`
`Output
`MCTF/DCVF -
`'
`'
`
`
`ecomposmon
`731
`
`
`
`
`Bitstream
`
`
`
`
`
`740
`
`Decoder
`
`D
`
`Side
`
`413
`
`Information
`
`En coder
`
`720
`
`FIG. 7
`
`741 \_/‘~'555
`
`
`
`1V8€l917l039002Sf]
`
`
`
`6110Llaaus900:“9'Inrnounanqnd"oueauddVelueIed
`
`
`
`
`
`
`
`
`
`Patent Application Publication
`
`Jul. 6, 2006 Sheet 8 of 19
`
`US 2006f0146138 A1
`
`
`
`ill-III"...
`
`III
`
`-u...
`
`
`
`
`
`Temporai Reference Pictures
`
`910
`
`Single-view RPL Manager
`
`FIG. 9
`
`PRIOR ART
`
`930
`
`
`
`
`Remove
`Insert
`Pictures
`Pictures
`
`
`
`
`901 [[01:13:31]anuoyeauddvwaned
` 6|106walls900:‘9“WT
`
`Ilafiflfiflflflal
`
` 960
`
`
`
`Temporal
`Prediction
`
`1V8€l917l039002Sf]
`
`
`
`
`1003
`J.
`Temcral Reference Pictures
`-
`
` l 00 1
`
` Multi-view
`
` RPL Manager
`
`Spatial Reference Pictures
`
`Synthesized Reference Pictures
`
`
`
`
`Insert
`
`Pictures
`
`Remove
`
`Pictures
`
`1020
`
`1030
`
`1040
`
`31050
`
`
`
`
`
`uoueauqnduoulzenddvmarred
`
`FIG. 10
`
`900E‘9‘[“I'
`6|.1001Balls
`IV8£l9|7l0l9003SI]
`
`1050\
`
`Multiview
`
`Prediction
`
`
`
`Patent Application Publication
`
`Jul. 6, 2006 Sheet 11 0f19
`
`US 2006f0146138 A1
`
`in?
`
`(0: -.—I
`yam;
`0-1.“):-
`E1383
`:m-me
`H" LI
`':§'£L
`
`-
`
`_ [iii-m]
`_ _-j-j (DH
`
`_- rug;
`
`1103
`
`1102
`
`1101
`
`FIG.11
`
`
`
`Patent Application Publication
`
`Jul. 6, 2006 Sheet 12 0f 19
`
`US 2006f0146138 Al
`
`:38IT
`
`com
`
`033:9.225
`
`NM65%
`
`
`
`Temporal Reference Pictures
`
`1003
`
`
`S . atial Reference Pictures
` Multi-view RPL
`
`View Mode
`100]
`Manager
`
`Synthesized Reference Pictures
`
`
`
`
`
`FIG. 13
`
`
`Multiview
`Prediction
`
`
`
`
`Reference Pictures
`
`with Higher Cor'relatio
`
`Reference Pictures
`Prediction Efficiency
`with Lower Correlation
`
`
`
`
`
`uoueauqnduouennddvJuamd
`
`900E‘9‘[“I'
`6|.10E1laallS
`IV8£l9|7l0l9003Si]
`
`
`
`._WB3MBatsu.u_
`.n1.u.nont4mnPnn23umzmnB..m.‘_.BmnB1t2wAI.nun..Bn._1t1umonmPUmlmmL0tom__Hu_
`
`__~.unuuuuen2nn1_Iw0u.H.mm0wn.mM_zn._1l__.nP.4u..._I......C_nuu..4"/\..unBm/\nBtsm_w
`
`v
`
`.
`
`
`
`
` .0011__7IIPPyn—v___ -_..wv..—a_.w_____nw0_nw_2.1z.4.n1nPn.w.n
`
`
`
`
`
`1-.--.....-.--.-.-.-..--.:11---:-----.::111:::--:.-:.
`
`tiJt6__5
`
`FIG. 14
`
`PRIOR ART
`
`
`
`
`
`3:2:>uumm32o=2232.5:._=_.93%mix:3on.o:m33.5333.5
`
`
`
`
`
`
`
`
`
`
`
`
`
`
`
` 6110§l133qS900:‘9‘Inrnonwlqndnoumuddwuaw
`
`
`
`[V85197[01'900ZSf]
`
`
`
`1905
`
`1930
`
`(//
`
`1903
`
`Reconstructed
`
`macroblock
`
`1920
`
`
`
`1904
`
`Synthesized
`macroblock
`
`1902
`
`View synthesis
`
`
`
`
`Decoded
`
`residual
`
`macroblock
`
`Spatial
`reference
`pictures
`
`
`
`4 1 3
`
` Side
`
`information
`
`decoder
`
`Depth
`
`1901
`
`FIG. 16
`
`uoynouqnduogeauddvwaned
`
`
`
`611091wallssum‘9'[fll‘
`IV8€l917l0f9002Sf]
`
`
`
`2020
`
`2010
`
`202 1
`
`2030
`
`Tem ral
`p0
`reference
`PICMBS
`
`
`2031
`.
`.
`
`
`Deterrnlne codlng
`.
`Motion
`cost usin tem oral
`
`
`
`estimation
`Motion
`prediction}: m 1
`
`vectors
`
`
`
`
`macroblock
`2042
`
`
`
`Determine coding
`cost using spatial
`
`
`pictures J
`Disparity
`
`
`prediction m
`vectors
`
`
`
`2070 Minimum
`C
`
`
`Determine coding
`
`
`Depth estimation
`
`cost using view-
`
`and View synthesis
`
`
`
`
`synthesis
`m3
`
`Synthesized
`
`View
`
`_____
`
`.
`‘
`I Spatia
`reference
`
`
`
`Determine coding
`_
`.
`cost using Intra
`redlction m4
`
`Neighboring
`pixels
`
`FIG. I 7
`
`209.0
`
`209 1
`
`Best
`
`mode
`
`
`
`uognailqnduogeanddvmated
`
`900z‘9'lM‘
`6l1“LIHalls
`IV8€l9l7l0l9003Sf]
`
`
`
`Decoded
`
`residual
`
`macroblock
`
`Spatial
`reference
`
`pictures
`
`
`
`
`. fl
`Synthesazed
`2121
`macroblock
`2120
`
`
`
`
`
`
`
`
`View synthesis
`
`2101
`
`Depth
`estimation
`
`Depth
`
`FIG. 18
`
`Reconstructed
`
`macroblock
`
`
`
`61108[was900:‘9“m“0933!!!!“(1uoueauddvmawcl
`
`[VsemwmoozSf]
`
`
`
`Patent Application Publication
`
`Jul. 6, 2006 Sheet 19 0f 19
`
`US 200610146138 A1
`
`
`
`
`
`
`
`
`
`.265%
`
`
`
`
`
`US 2006110146138 Al
`
`Jul. 6, 2006
`
`METHOD AND SYSTEM FOR SYN'lil-IESIZING
`MULTIVIEW VIDEOS
`
`REIAII‘IEI) APPLICATION
`
`[0001] This application is a continuation—in—part of US.
`patent application Ser. No. 111015390 entitled “Multiview
`Video Decomposition and Encoding” and filed by Xin et al.
`on Dec. 17, 2004. This application is related to US patent
`application Ser. No. XXIXXXJXX entitled “Method and
`System for Managing Reference Pictures in Multiview Vid—
`eos“ and U.S. patent application Ser. No. XXIXXX,XXX
`entitled “Method for Randomly Accessing Mulliview Vid-
`eos“, both of which were co-filed with this application by
`Xin et al. on Nov. 30, 2005.
`
`I:ll'.l.l.,l) ()l" Tllli INV’ltNTlON
`
`[0002] This invention relates generally to encoding and
`decoding multiview videos, and more particularly to syn—
`thesizing tnultiview videos.
`
`BACKGROUND OF THE INVENTION
`
`[0003] Multivimv video encoding and decoding is essen-
`tial for applications such as three dimensional television
`(3DTV), free viewpoint television (FTV), and multicamera
`surveillance. Multiview video encoding and decoding is also
`known as dynamic light field compression.
`
`[0004] FIG. 1 shows a prior art ‘simulcast‘ system 100 for
`multiview video encoding. Cameras 1-4 acquire sequences
`of frames or videos 101-104 of a scene 5. Each camera has
`a different view ofthe scene. Each video is encoded 111—114
`independently to corresponding encoded videos 121—124.
`That system uses conventional 2D video encoding tech—
`niques. Therefore, that system does not correlate between
`the different videos acquired by the cameras from the
`difl'erent viewpoints while predicting frames of the encoded
`video. Independent encoding decreases compression effi—
`ciency, and thus network bandwidth and storage are
`increased.
`
`[0005] FIG. 2 shows a prior art disparity compensated
`prediction system 200 that does use inter-view correlations.
`Videos 201—204 are encoded 211—214 to encoded videos
`
`231—234. The videos 201 and 204 are encoded independently
`using a standard video encoder such as MPEG—2 or H.264,
`also known as MPEG-4 Part 10. These independently
`encoded videos are ‘reference’ videos. The remaining videos
`202 and 203 are encoded using temporal prediction and
`inter—view predictions based on reconstructed reference vid—
`eos 251 and 252 obtained from decoders 221 and 222.
`
`Typically, the prediction is determined adaptively on a per
`block basis. S. C. Chan et al., “The data compression of
`simplified dynamic light fields.“ Proc. IEEE Int. Acoustics,
`Speech, and Signal Processing Conf.. April, 2003.
`
`‘lifting—based’ wavelet
`[0006] FIG. 3 shows prior art
`decomposition. see W. Sweldens, “The data compression of
`simplified dynamic light fields,” .1. App]. Comp.
`llarm.
`Anal, vol. 3, no. 2. pp. 186-200, 1996. Wavelet decompo-
`sition is an effective technique for static light field compres—
`sion. Input samples 301 are split 310 into odd samples 302
`and even samples 303. The odd samples are predicted 320
`from the even samples. A prediction error fomts high band
`samples 304. The high band samples are used to update 330
`
`the even samples and to form low band samples 305. That
`decomposition is invertible so that
`linear or non—linear
`operations can be incorporated into the prediction and
`update steps.
`
`[0007] The lifting scheme enables a motion—compensated
`temporal transfonn, i.e., motion compensated temporal fil—
`tering (MCTF) which, for videos. essentially filters along a
`temporal motion trajectory. A review of MCTF for video
`coding is described by Ohm et al., “Interframe wavelet
`coding—motion picture representation for universal scal—
`ability,“ Signal Processing: Image Communication, vol. 19.
`no. 9, pp. 877-908. October 2004. The lifting scheme can be
`based on any wavelet kernel such as Man or 523 Daubechies,
`and any motion model such as block—based translation or
`affine global motion, without affecting the reconstruction.
`
`[0008] For encoding, the MCTF decomposes the video
`into high band frames and low band frames. Then,
`the
`frames are subjected to spatial transforms to reduce any
`remaining spatial correlations. The transformed low and
`high band frames, along with associated motion information.
`are entropy encoded to form an encoded bitstream. MCTF
`can be implemented using the lifting scheme shown in FIG.
`3 with the temporally adjacent videos as input. In addition,
`MCTF can be applied recursively to the output low band
`frames.
`
`[0009] MC'I‘li-based videos have a compression efficiency
`comparable to that of video compression standards such as
`H.2641’AVC. In addition, the videos have inherent temporal
`scalability. However, that method cannot be used for directly
`encoding multiview videos in which there is a correlation
`between videos acquired from multiple views because there
`is no efficient method for predicting views that accounts for
`correlation in time.
`
`[0010] The lifting scheme has also been used to encode
`static light fields, i.e., single multiview images. Rather than
`performing a motion—compensated temporal filtering, the
`encoder performs a disparity compensated inter—view filter—
`ing (DEW/'1’) across the static views in the spatial domain. see
`Chang et al.. “Inter-view wavelet compression of light fields
`with disparity compensated lifting,” SPIE Conf on Visual
`Communications and Image Processing, 2003. For encod—
`ing, DCVF decomposes the static light field into high and
`low band images, which are then subject to spatial trans-
`forms to reduce any remaining spatial correlations. The
`transformed images, along with the associated disparity
`information, are entropy encoded to form the encoded
`bitstream. [JCVI’ is typically implemented using the lifting-
`bascd wavelet transform scheme as shown in FIG. 3 with
`the images acquired from spatially adjacent camera views as
`input. In addition, DCVF can be applied recursively to the
`output
`low band images. DCVF—based static light
`field
`compression provides a better compression efficiency than
`independently coding the multiple frames. However, that
`method also cannot encode multiview videos in which both
`
`temporal correlation and spatial correlation between views
`are used because there is no efficient method for predicting
`views that account for correlation in time.
`
`SUMMARY OF THE INVENTION
`
`[0011] A method and system to decompose multiview
`videos acquired ofa Scene by multiple cameras is presented.
`
`
`
`US 2006f0146l38 A1
`
`Jul. 6, 2006
`
`[0012] Each multiview video inclttdes a sequence of
`frames, and each camera provides a different view of the
`scene.
`
`[0027] FIG. 1] is a block diagram of multiview reference
`pictures in a decoded picture bufl'er according to an embodi—
`ment of the invention;
`
`[0013] A prediction mode is selected from a temporal,
`spatial, view synthesis. and intra—prediction mode.
`
`[0028] FIG. 12 is a graph comparing coding efliciencies of
`different multiview reference picture orderings;
`
`[0014] The multiview videos are then decomposed into
`low band frames. high band frames, and side information
`according to the selected prediction mode.
`
`[0015] A novel video reflecting a synthetic view of the
`scene can also be generated from one or more of the
`multiview videos.
`
`[0016] More particularly. one embodiment of the invert—
`tion provides a system and method for synthesizing multi—
`view videos. Multiview videos are acquired of a scene with
`corresponding cameras arranged at a poses such that there is
`view overlap between any pair of cameras. A synthesized
`multiview video is generated from the acquired multiview
`videos for a virtual camera. A reference picture list
`is
`maintained for each current frame of each of the multiview
`videos and the synthesized video. The reference picture list
`indexes temporal reference pictures and spatial reference
`pictures of the acquired multiview videos and the synthe—
`sized reference pictures of the synthesized multiview video.
`Then, each current frame of the multiview videos is pre-
`dicted according to reference pictures indexed by the asso—
`ciated reference picture list during encoding and decoding.
`
`BRIEF DESCRIPTION 01" THE DRAWINGS
`
`[0017] FIG. 1 is a block diagram of a prior art system for
`encoding multiview videos;
`
`[0018] FIG. 2 is a block diagram ofa prior art disparity
`compensated prediction system for encoding multiview vid—
`eos;
`
`[0019] FIG. 3 is a flow diagram of a prior art wavelet
`decomposition process;
`
`[0020] FIG. 4 is a block diagram of a MCTFJDCVF
`decomposition according to an embodiment of the invert—
`tion;
`
`[0021] FIG. 5 is a block diagram of low—band frames and
`high band frames as a function of time and space after the
`MC'I‘WIXIVF decomposition according to an embodiment
`of the invention:
`
`[0022] FIG. 6 is a block diagram of prediction of high
`band frame from adjacent low-band frames according to an
`embodiment of the invention:
`
`[0023] FIG. 7 is a block diagram of a multiview coding
`system using macroblock-adaptive MC’I‘FI’I.)(TVF decompo-
`sition according to an embodiment of the invention;
`
`[0024] FIG. 8 is a schematic of video synthesis according
`to an embodiment of the invention;
`
`[0025] FIG. 9 is a block diagram ofa prior an reference
`picture management;
`
`[0026] FIG. 10 is a block diagram of multiview reference
`picture management according to an embodiment of the
`invention:
`
`[0029] FIG. 13 is a block diagram of dependencies of
`view mode on the multiview reference picture list manager
`according to an embodiment of the invention;
`
`[0030] FIG. 14 is diagram ofa prior art reference picture
`management for single view coding systems that employ
`prediction from temporal reference pictures;
`
`[0031] FIG. 15 is a diagram of a reference picture man-
`agement for multiview coding and decoding systems that
`employ prediction from multiview reference pictures
`according to an embodiment of the invention;
`
`[0032] FIG. 16 is a block diagram of view synthesis in a
`decoder using depth information encoded and received as
`side information according to an embodiment of the inven-
`11(3t11
`
`[0033] FIG. 17 is a block diagram of cost calculations for
`selecting a prediction mode according to an embodiment of
`the invention;
`
`[0034] FIG. 18 is a block diagram of view synthesis in a
`decoder using depth imomtation estimated by a decoder
`according to an embodiment of the invention; and
`
`[0035] FIG. 19 is a block diagram of multiview videos
`using V—frames to achieve spatial random access in the
`decoder according to an embodiment of the invention.
`
`[)I:.".'IL’\II..EI) DESCRIPTION 01" 'l‘IIIi
`EMBODIMENTS OF THE INVENTION
`
`[0036] One embodiment of our invention provides a joint
`temporalfinter—view processing method for encoding and
`decoding frames of multiview videos. Multiview videos are
`videos that are acquired of a scene by multiple cameras
`having different poses. We define a pose camera as both its
`3]) (x, y. 7.) position. and its 3D (0. p, 1p) orientation. Each
`pose corresponds to a ‘view’ of the scene.
`
`[0037] The method uses temporal correlation between
`frames within each video acquired for a particular camera
`pose, as well as spatial correlation between synchronized
`frames in videos acquired from multiple camera vieWs. In
`addition, ‘synthetic’ frames can be correlated, as deseribed
`below.
`
`In one embodiment, the temporal correlation uses
`[0038]
`motion compensated temporal filtering (MCTF), while the
`spatial correlation uses disparity compensated inter—view
`filtering [DCVI’).
`
`In another embodiment of the invention, spatial
`[0039]
`correlation uses prediction of one view from synthesized
`frames that are generated from ‘neighboring’ frames. Neigh-
`boring frames are temporally or spatially adjacent frames,
`for example, frames before or after a current frame in the
`temporal domain, or one or more frames acquired at the
`same instant in time but from cameras having different poses
`or vievvs of the scene.
`
`liach frame of each video includes macrobhicks of
`[0040]
`pixels. Therefore. the method of multiview video encoding
`
`
`
`US 2006r’0146l38 Al
`
`Jul. 6, 2006
`
`and decoding according to one embodiment ofthe invention
`is macroblock adaptive. The encoding and decoding of a
`current macroblock in a current frame is performed using
`several possible prediction modes, including various forms
`of temporal, spatial. view synthesis. and intra prediction. To
`determine the best prediction mode on a macroblock basis,
`one embodiment of the invention provides a method for
`selecting a prediction mode. The method can be used for any
`number of camera arrangements.
`
`In order to maintain compatibility with existing
`[0041]
`single-view encoding and decoding systems, a method for
`managing a reference picture list is described. Specifically,
`we describe a method of inserting and removing reference
`pictures from a picture bufl'er according to the reference
`picture list. 'lhe reference pictures include temporal refer-
`ence pictures, spatial reference pictures and synthesized
`reference pictures.
`
`[0042] As used herein, a reference picture is defined as any
`frame that
`is used during the encoding and decoding to
`‘predict‘ a current frame. Typically. reference pictures are
`spatially or temporally adjacent or ‘neighboring‘ to the
`current frame.
`
`It is important to note that the same operations are
`[0043]
`applied in both the encoder and decoder because the same
`set of reference pictures are used at tmy give time instant to
`encode and decode the current frame.
`
`[0044] One embodiment of the invention enables random
`access to the frames of the multiview videos during encod—
`ing and decoding. This improves coding efliciency.
`
`[0045] MCTFIDCVF Decomposition
`
`[0046] FIG. 4 show a MCI‘FI‘IITVF decomposition 400
`according to one embodiment of the invention. Frames of
`input videos 401-404 are acquired of a scene 5 by cameras
`1-4 having different posses. Note. as shown in FIG. 8, some
`of the cameras la and lb can be at the same locations but
`with different orientations. It is assumed that there is some
`amount of view overlap between any pair of cameras. The
`poses of the cameras can change while acquiring the mul—
`tiview videos. Typically, the cameras are synchronived with
`each other. l'iach input video provides a different ‘view’ of
`the scene. The input frames 401—404 are sent to a MC‘TFtr
`DCVF decomposition 400. The decomposition produces
`encoded low—band frames 411, encoded high band frames
`4] 2, and associated side infonnation 413. The high band
`frames encode prediction errors using the low band frames
`as reference pictures. The decomposition is according to
`selected prediction modes 410. The prediction modes
`include spatial. temporal. view synthesis. and intra predic-
`tion modes. The prediction modes can be selected adaptively
`on a per macroblock basis for each current frame. With intra
`prediction, the current macroblock is predicted from other
`macroblocks in the same frame.
`
`[0047] FIG. 5 shows a preferred alternating ‘checkerboard
`pattern’ of the low band frames (L) 411 and the high band
`frames (II) 412 for a neighborhood of frames 510. The
`frames have a spatial (view) dimension 501 and a temporal
`dimension 502. Essentially, the pattern alternates low band
`frames and high band frames in the spatial dimension for a
`single instant in time, and additionally alternates temporally
`the low band frames and the high band frames for a single
`video.
`
`[0048] There are several advantages of this checkerboard
`pattern. The pattern distributes low band frames evenly in
`both the space and time dimensions, which achieves scal—
`ability in space and time when a decoder only reconstructs
`the low baud frames. In addition, the pattern aligns the high
`band frames with adjacent low band frames in both the space
`and time dimensions. This maximizes the correlation
`
`between reference pictures from which the predictions ofthe
`errors in the current frame are made, as shown in FIG. 6.
`
`[0049] According to a lifting—based wavelet transfoml. the
`high band frames 412 are generated by predicting one set of
`samples from the other set of samples. The prediction can be
`achieved using a number of modes including various forms
`of temporal prediction. various forms of spatial prediction,
`and a view synthesis prediction according to the embodi—
`ments of invention described below.
`
`[0050] The means by which the high band frames 412 are
`predicted and the necessary information required to make
`the prediction are referred to as the side infonnation 413. If
`a temporal prediction is performed, then the temporal mode
`is signaled as part of the side information along with
`corresponding motion information. Ifa spatial prediction is
`performed, then the spatial mode is signaled as part of the
`side information along with corresponding disparity infor-
`mation. If view synthesis prediction is performed, then the
`view synthesis mode is signaled as part of the side infor—
`lnation along with corresponding diSparity. motion and
`depth in formation.
`
`the prediction of each
`[0051] As shown in FIGS. 6,
`current frame 600 uses neighboring frames 510 in both the
`space and time dimensions. The frames that are used for
`predicting the current frame are called reference pictures.
`The reference pictures are maintained in the reference list,
`which is part of the encoded bitstream. The reference
`pictures are stored in the decoded picture buffer.
`
`In one embodiment of the invention, the MC’I‘I"
`[0052]
`aud DCVF are applied adaptively to each current macrob-
`lock for each frame of the input videos to yield decomposed
`low band frames, as well as the high band frames and the
`associated side infonnation. In this way, each macroblock is
`processed adaptively according to a ‘best‘ prediction mode.
`An optimal method for selecting the prediction mode is
`described below.
`
`In one embodiment of the invention, the MCTI" is
`[0053]
`first applied to the frames of each video independently. The
`resulting frames are then further decomposed with the
`DCVF. In addition to the final decomposed frames,
`the
`corresponding side information is also generated. lf per-
`formed on a macroblock-basis, then the prediction mode
`selections for the MCTF and the DCVF are considered
`
`separately. As an advantage, this prediction mode selection
`inherently supports temporal scalability. In this way, lower
`temporal rates of the videos are easily accessed in the
`compressed bitstream.
`
`In another embodiment, the [XIV]? is first applied
`[0054]
`to the frames of the input videos. The resulting frames are
`then temporally decomposed with the MCTF. In addition to
`the final decomposed frames, the side information is also
`generated. If performed on a macroblock—basis, then the
`prediction mode selections for the MC’I‘F and DCVIJ are
`considered separately. As an advantage, this selection inher-
`
`
`
`US 200670146138 Al
`
`Jul. 6, 2006
`
`ently supports spatial scalability. In this way. a reduced
`number of the views are easily accessed in the compressed
`bitstream.
`
`[0055] The decomposition described above can be applied
`recursively on the resulting set of low band frames from a
`previous decomposition stage. As an advantage, our MC‘TFtr
`DCVF decomposition 400 effectively removes both tempo—
`ral and spatial (inter-view] correlations, and can achieve a
`very high compression efficiency. The compression elli-
`cieney of our multiview video encoder outperforms conven-
`tional simulcast encoding. which encodes each video for
`each view independently.
`
`[0056] Coding of MC'l‘Ft'DCVI: Decomposition
`
`[0057] As shown in FIG. 7. the outputs 411 and 412 of
`decomposition 400 are fed to a signal encoder 710, and the
`output 413 is fed to a side information encoder 720. The
`signal encoder 710 performs a transfonn, quantization and
`entropy coding to remove any remaining correlations in the
`decomposed low band and high band frames 411—412. Such
`operations are well known in the art. Netravali and Ilaskell,
`Digital Pictures: Representation. Compression and Stan-
`dards, Second lfldition, Plenum Press. 1995.
`
`[0058] The side information encoder 720 encodes the side
`information 413 generated by the decomposition 400. In
`addition to the prediction mode and the reference picture list,
`the side information 413 includes motion information cor—
`
`responding to the temporal predictions, disparity informa—
`tion corresponding to the spatial predictions and view syn-
`thesis and depth information corresponding to the view
`synthesis predictions.
`
`tincoding the side information can be achieved by
`[0059]
`known and established techniques, such as the techniques
`used in the MPEG—4 Visual standard, ISOIIEC 14496—2,
`“Information
`technology—Coding
`of
`audio—visual
`objects
`Part 2: Visual,” 2"" Edition. 2001, or the more
`recent [1.264I'AVC standard, and [TU-T Recommendation
`H.264, “Advanced video coding for generic audiovisual
`services,” 2004.
`
`[0060] For instance. motion vectors of the macroblocks
`are typically encoded using predictive methods that deter—
`mine a prediction vector from vectors in macroblocks in
`reference pictures. The difference between the prediction
`vector and the current vector is then subject to an entropy
`coding process, which typically uses the statistics of the
`prediction error. A similar procedure can be used to encode
`disparity vectors.
`
`[0061] Furthermore, depth information for each macrob—
`1ock can be encoded using predictive coding methods in
`which a prediction from macroblocks in reference pictures is
`obtained, or by simply using a fixed length code to express
`the depth value directly. If pixel level accuracy for the depth
`is extracted and compressed, then texture coding techniques
`that apply transform, quantization and entropy coding tech
`niques can be applied.
`
`[0062] The encoded signals 711—713 from the signal
`encoder 710 and side information encoder 720 can be
`multiplexed 730 to produce an encoded output bitstream
`731.
`
`[0063]
`
`Decoding of M(f'l‘lifl.)(.‘VF I.)ecomposition
`
`[0064] The bitstream 731 can be decoded 740 to produce
`output multiview videos 741 corresponding to the input
`multiview videos 401—404. Optionally, synthetic video can
`also be generated. Generally,
`the decoder performs the
`inverse operations of the encoder to reconstruct the multi-
`view videos. If all
`low band and high band frames are
`decoded, then the full set of frames in both the space (view)
`dimension and time dimension at the encoded quality are
`reconstructed and available.
`
`[0065] Depending on the number of recursive levels of
`decomposition that were applied in the encoder and which
`type of decompositions were applied. a reduced number of
`videos andfor a reduced temporal rate can be decoded as
`shown in FIG. 7.
`
`[0066] View Synthesis
`
`[0067] As shown in FIG. 8, view synthesis is a process by
`which frames 801 ofa synthesized video are generated from
`frames 803 of one or more actual mu ltiview videos. In other
`
`words. view synthesis provides a means to synthesize the
`frames 80] corresponding to a selected novel view 802 of
`the scene 5. This novel view 802 may correspond to a
`‘virtual’ camera 800 not present at
`the time the input
`multiview videos 401—404 were acquired or the view can
`correspond to a camera view that is acquired. whereby the
`synthesized View will he used for prediction and encoding.If
`decoding of this view as described below.
`
`Ifone video is used. then the synthesis is based on
`[0068]
`extrapolation or warping, and if multiple videos are used.
`then the synthesis is based on interpolation.
`
`[0069] Given the pixel values of frames 803 of one or
`more multiview videos and the depth values of points in the
`scene, the pixels in the frames 80] for the synthetic view 802
`can be synthesized from the corresponding pixel values in
`the frames 803.
`
`[0070] View synthesis is commonly used in computer
`graphics for rendering still images for multiple views. see
`Buehler et al.. “Unstructured Lumigraph Rendering,“ Proc.
`ACM SIGGRAPH, 2001. That method requires extrinsic
`and intrinsic parameters for the cameras.
`
`[0071] View synthesis for compressing multiview videos
`is novel. In one embodiment of our invention. we generate
`synthesized frames to be used for predicting the current
`frame. In one embodiment of the invention, synthesized
`frames are generated for designated high band frames. In
`another embodiment of the invention, synthesized frames
`are generated for specific views. The synthesized frames
`serve as reference pictures from which a current synthesized
`frame can be predicted.
`
`[0072] One difiziculty with this approach is that the depth
`values of the scene 5 are unknown. Therefore, we estimate
`the depth values using known techniques, e.g., based on
`correspondences of features in the multiview videos.
`
`[0073] Alternatively, for each synthesized video, we gen-
`erate multiple synthesized frames, each corresponding to a
`candidate depth value. For each macroblock in the current
`frame, the best matching macroblock in the set of synthe—
`sized frames is determined. The synthesized frame from
`which this best match is found indicates the depth value of
`
`
`
`US 2006f0146l38 Al
`
`Jul. 6, 2006
`
`the macroblock in the current frame. This process is repeated
`for each macroblock in the current frame.
`
`[0074] A difference between the current macroblock and
`the synthesized block is encoded and compressed by the
`signal encoder 710. The side information for this multiview
`mode is encoded by the side information encoder 720. The
`side infonnation includes a signal indicating the view syn-
`thesis prediction mode, the depth value of the macroblock,
`and an optional displacement vector that compensates for
`any misaligmnents between the macroblock in the current
`frame and the best matching macmblock in the synthesized
`frame to be compensated.
`
`[0075] Prediction Mode Selection
`
`In the macroblock—adaptive MCTFfDCVF decom—
`[0076]
`position, the predict'ion mode in for each macroblock can be
`selected by minimizing a cost function adaptively on a per
`macroblock basis:
`
`m*=mg min J(m},
`
`where J(m)=l)(m)+?&R(m). and I) is distortion, A is a weight-
`ing parameter, R is rate, m indicates the set of candidate
`prediction modes, and m‘“ indicates the optimal prediction
`mode that has been selected based on a minimum cost
`criteria.
`
`[0077] The candidate modes In inclu