[go: up one dir, main page]

WO2006108863A2 - Process for scalable coding of images - Google Patents

Process for scalable coding of images Download PDF

Info

Publication number
WO2006108863A2
WO2006108863A2 PCT/EP2006/061563 EP2006061563W WO2006108863A2 WO 2006108863 A2 WO2006108863 A2 WO 2006108863A2 EP 2006061563 W EP2006061563 W EP 2006061563W WO 2006108863 A2 WO2006108863 A2 WO 2006108863A2
Authority
WO
WIPO (PCT)
Prior art keywords
layer
macroblock
base
block
blocks
Prior art date
Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
Ceased
Application number
PCT/EP2006/061563
Other languages
French (fr)
Inventor
Edouard Francois
Jérome Vieron
Gwenaelle Marquant
Nicolas Burdin
Patrick Lopez
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Thomson Licensing SAS
Original Assignee
Thomson Licensing SAS
Priority date (The priority date is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the date listed.)
Filing date
Publication date
Application filed by Thomson Licensing SAS filed Critical Thomson Licensing SAS
Publication of WO2006108863A2 publication Critical patent/WO2006108863A2/en
Anticipated expiration legal-status Critical
Ceased legal-status Critical Current

Links

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
    • H04N19/59—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving spatial sub-sampling or interpolation, e.g. alteration of picture size or resolution
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
    • H04N19/587—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving temporal sub-sampling or interpolation, e.g. decimation or subsequent interpolation of pictures in a video sequence

Definitions

  • the invention relates to a process for coding/decoding of images, more specifically scalable video coding/decoding where several spatial video formats need to be encoded/decoded.
  • This process enables the encoding of two spatial resolutions, low and high layers, where the factor between pictures width and height of the two successive spatial layers can be different vertically and horizontally and does not necessarily equal 2 and where pictures of the higher resolution level can contain parts, for example picture borders, that are not present in corresponding pictures of the low resolution level.
  • This process is fully generic, that is, any inter-layer size ratio and any cropping window can be handled. It addresses both the "spatial scalable coding using cropped areas” (CE9) and “non dyadic scalability” (CE10).
  • Figure 1 shows relations between the enhancement layer 1 and the base layer 2.
  • Width and height of enhancement layer pictures are defined respectively as w enh and h enh .
  • Base layer pictures dimensions are defined as w base and h base .
  • Base layer pictures are a subsampled version of sub-pictures of enhancement layer pictures, of dimensions w extracl and h extracl , positioned at coordinates (x O ⁇ g , Y o rig) in the enhancement layer pictures coordinates system.
  • the enhancement and base layer pictures are divided in macroblocks.
  • Parameters [X 0 O 9 , / «& , w extrac t, hextmct) completely define the geometrical relations between high and low layer pictures. In a standard dyadic scalable version, these parameters are equal to (0,0,2. w base ,2.h extracl ).
  • Figure 2 illustrates the macroblock overlapping between the upsampled base layer picture 3 , in dashed lines, and the enhancement layer picture, in solid lines.
  • a macroblock of enhancement layer may have either no base layer corresponding block, on borders of the enhancement layer picture, either one or several base layer corresponding macroblocks. Consequently a specific managing of the inter layer prediction is necessary.
  • An aim of the invention is to solve the aforementioned problems.
  • Its object is a process for scalable coding of images with a F1 format corresponding to an enhanced layer and with a F2 format having a lower resolution than F1 , corresponding to a base layer, comprising an inter layer residual prediction mode to code a current macroblock of the enhancement layer from a prediction macroblock of the base layer, a macroblock being made up of elementary blocks, the prediction macroblock being obtained through an upsampling of the base layer, characterised in that the upsampling is made at an elementary block level and in that the phase of the top left sample within an elementary block depends on the location of the elementary block within the picture.
  • the size of an elementary block is 4 x 4 pixels.
  • the filter used for the upsampling is a 6-tap filter.
  • the upsampling is a 4/3 upsampling corresponding to 4 samples for 3 original pixels.
  • the phase expressed as the relative position of the pixel, for 3 successive elementary blocks, is 0, ⁇ A and 1 A
  • a high layer macroblock can exploit inter-layer prediction using scaled base layer motion data, using either "BASE_LAYER_MODE” or “QPEL_REFINEMENT_MODE", as in case of dyadic spatial scalability.
  • the high layer macroblock is reconstructed with default motion data deduced from the base layer.
  • These modes are only authorized for high layer macroblocks having corresponding base layer blocks, i.e. grey-colored area in figure 3 where bold line represents the upsampled base layer window 3.
  • these macrob lock modes indicate that motion/prediction information including macroblock partitioning are directly derived from the base layer.
  • the process consists in constructing a prediction macroblock, MB pred, inheriting motion data from base layer.
  • the macroblock partitioning as well as the reference indices and motion vectors are those of the prediction macroblock MBjpred.
  • QPEL_REFINEMENT_MODE is similar, but a quarter-sample motion vector refinement is achieved.
  • the inheriting process consists in considering the center point (x c , y c ) of the block, computing its base layer luma location (xB, yB), and inheriting motion data from the base layer block containing the pixel (xB, yB).
  • Figure 4 represents the localisation of a high layer pixel in the base layer.
  • (x ⁇ ,y ⁇ ) be the luma coordinates of the upper left pixel of b.
  • the base luma location ( xB, yB ) of the pixel (1 ,1 ) of b in the base layer is derived as follows :
  • BasePicWidthlnSamples L 16 * ( BasePicWidthlnMbsMinus1+1 )
  • BasePicHeightlnSamples L 16*( BasePicHeightlnMbsMinus1+1 )
  • the base corresponding macroblock is defined as the base layer macroblock containing pixel ( xB, yB ). Its coordinates ( mbxB, mbyB ) are given by the following equations:
  • the base corresponding 4x4 block is defined as the base layer 4x4 block containing pixel ( xB, yB ). Its coordinates ( b4xB, b4yB ) in the base corresponding macroblock are defined by the following equations:
  • the 4x4 block gets the reference index and motion vector from the base corresponding 4x4 block
  • reference indices and partitioning mode are chosen as follows:
  • BLK_8x8 is chosen • else BLK_8x4 is chosen
  • a final process is achieved to determine the MB pred mode.
  • 8x8 blocks of the macroblock are identified as indicated in figure 6, blocks B1 and B2 for the top, B3 and B4 for the bottom, from the left to the right.
  • Two 8x8 blocks are considered as identical blocks if: - One or both of the two 8x8 blocks are classified as intra blocks or
  • Partitioning mode of both blocks is BLK_8x8 and reference indices and motion vectors of listO and Iist1 of each 8x8 block, if they exist, are identical.
  • the mode choice is done using the following process:
  • MB pred is an INTER macroblock Remaining intra 8x8 blocks are enforced to be Inter 8x8 blocks o 8x8 blocks classified as intra are enforced to INTER blocks with 8x8 partitioning. Their reference indices and motion vectors are computed as follows. Let B, NT RA be such a 8x8 block. for each list listx
  • motion vector rescaling is finally applied to every existing motion vectors of the prediction macroblock MB pred.
  • Inter layer texture prediction is based on the same principles as inter layer motion prediction. It is only possible for macroblocks fully embedded in the scaled base layer window (grey-colored area in figure 3). For Intra texture prediction, the interpolation filter is applied across transform blocks boundaries. For residual texture prediction, this process only works inside transform blocks (4x4 or 8x8 depending on the transform).
  • the base layer prediction array corresponds to the samples contained in the area (xB-8, yB-8) and (xB+16, yB+16).
  • the same filling process, as used in the dyadic case, is applied to fill samples areas corresponding to non existing or non available samples (for instance, in case of intra texture prediction, samples that do not belong to intra blocks).
  • the base layer prediction array is then upsampled.
  • the upsampling is applied in two steps : first, texture is upsampled using the AVC half pixel 6-tap filter; then a bilinear interpolation is achieved to build the quarter pel samples, which results in a quarter-pel interpolation array. For intra texture, this interpolation crosses block boundaries. For residual texture, interpolation does not cross transform block boundaries.
  • a given macroblock MB of current layer can exploit intra layer residual prediction only if co-located macroblocks of the base layer exist and are intra macroblocks.
  • the corresponding 8x8 blocks of the base layer high-pass signal are directly de-blocked and interpolated, as in case of 'standard' dyadic spatial scalability.
  • the same padding process is applied for deblocking. - inter-layer residual prediction
  • a given macroblock MB of current layer can exploit inter layer residual prediction only if co-located macroblocks of the base layer exist and are not intra macroblocks.
  • the upsampling process consists in upsampling each elementary transform block, without crossing the block boundaries. For instance, if a MB is coded into four 8x8 blocks, four upsampling processes will be applied on exactly 8x8 pixels as input.
  • the interpolation process is achieved in two steps : first, the base layer texture is upsampled using the AVC half pixel 6-tap filter; then a bilinear interpolation, is achieved to build the quarter pel samples. Interpolated high layer samples The nearest quarter pel position is chosen as the interpolated pixel.
  • Figure 7 shows how the 4x4 blocks are oversampled in the case of a 4/3 ratio.
  • the number of pixels per block in the original picture is 4x4. After the 4/3 oversampling, this number becomes 6x6, 6x5, 5x6 or 5x5 depending on the location of the block within the picture. It can be observed that this number of pixels inside an oversampled 4x4 block follows a regular periodic pattern. In the 4/3 case, the pattern is : ⁇ 6, 5, 5 ⁇
  • phase of the top left pixel of each expressed as the position of this pixel relatively to original grid is given by another periodic pattern.
  • the pattern is : ⁇ 0, 1/2, 1/4 ⁇ .

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)
  • Compression, Expansion, Code Conversion, And Decoders (AREA)

Abstract

The process, comprising an inter layer residual prediction mode, to code a current macroblock of an en hancement layer from a prediction macroblock of a base layer, a macroblock being made up of elementary blocks, the prediction macroblock being obtained through an upsampling of the base layer (3) is characterised in that the upsampling is made at an elementary block level and in that the phase of the top left sample within an elementary block depends on the location of the elementary block within the picture.

Description

Process for scalable coding of images
The invention relates to a process for coding/decoding of images, more specifically scalable video coding/decoding where several spatial video formats need to be encoded/decoded.
This process enables the encoding of two spatial resolutions, low and high layers, where the factor between pictures width and height of the two successive spatial layers can be different vertically and horizontally and does not necessarily equal 2 and where pictures of the higher resolution level can contain parts, for example picture borders, that are not present in corresponding pictures of the low resolution level.
This process is fully generic, that is, any inter-layer size ratio and any cropping window can be handled. It addresses both the "spatial scalable coding using cropped areas" (CE9) and "non dyadic scalability" (CE10).
We consider two successive spatial layers, a low layer, considered as base layer, and a high layer, considered as enhancement layer. Figure 1 shows relations between the enhancement layer 1 and the base layer 2. Width and height of enhancement layer pictures are defined respectively as wenh and henh. Base layer pictures dimensions are defined as wbase and hbase. Base layer pictures are a subsampled version of sub-pictures of enhancement layer pictures, of dimensions wextracl and hextracl, positioned at coordinates (xOπg , Yorig) in the enhancement layer pictures coordinates system. In figure 1 , the enhancement and base layer pictures are divided in macroblocks. Upsampling factors between base layer pictures and extraction pictures in enhancement layer are respectively defined as αhOήz = weχtract I wbase and αvertiC = ^extract I hbase for horizontal and vertical dimensions. Parameters [X0O9 , /«& , wextract, hextmct) completely define the geometrical relations between high and low layer pictures. In a standard dyadic scalable version, these parameters are equal to (0,0,2. wbase,2.hextracl).
The problem to solve is the encoding/decoding of the macroblocks of the enhancement layer knowing the decoded base layer. Figure 2 illustrates the macroblock overlapping between the upsampled base layer picture 3 , in dashed lines, and the enhancement layer picture, in solid lines. A macroblock of enhancement layer may have either no base layer corresponding block, on borders of the enhancement layer picture, either one or several base layer corresponding macroblocks. Consequently a specific managing of the inter layer prediction is necessary.
An aim of the invention is to solve the aforementioned problems.
Its object is a process for scalable coding of images with a F1 format corresponding to an enhanced layer and with a F2 format having a lower resolution than F1 , corresponding to a base layer, comprising an inter layer residual prediction mode to code a current macroblock of the enhancement layer from a prediction macroblock of the base layer, a macroblock being made up of elementary blocks, the prediction macroblock being obtained through an upsampling of the base layer, characterised in that the upsampling is made at an elementary block level and in that the phase of the top left sample within an elementary block depends on the location of the elementary block within the picture.
According to a particular embodiment, the size of an elementary block is 4 x 4 pixels.
According to a particular embodiment, the filter used for the upsampling is a 6-tap filter. According to a particular embodiment, the upsampling is a 4/3 upsampling corresponding to 4 samples for 3 original pixels.
According to a particular embodiment, the phase, expressed as the relative position of the pixel, for 3 successive elementary blocks, is 0, ΛA and 1A
Other objects and features of the invention will become understood from the following description with reference to the accompanying drawings which represent :
- figure 1 , relations between the enhancement layer and base layer,
- figure 2, macroblock overlapping between upsampled base layer picture and enhancement layer picture,
- figure 3, authorized blocks for inter layer motion vector prediction,
- figure 4, localization of a high layer pixel in the base layer,
- figure 5, 4x4 blocks b of a 8x8 block B
- figure 6, 8x8 blocks B of a prediction macroblock, - figure 7, oversampling of the 4x4 blocks in the case of a 4/3 ratio.
In the sequel, according to the semantic defined further, we will name: - scaled_base_column = xorig
- scaled_base_line = yOπg
- scaled_base_width = wextract
- scaled_base_height = hextracl
Inter layer prediction is only possible for macroblocks fully embedded in the scaled base layer window, corresponding to the grey-colored area in figure 3, that is, macroblocks whose coordinates (MBχ,Mby) respect the following conditions: MBx >= (scaled_base_column + 15) / 16 and
MBx < (scaled_base_column + scaled_base_width) / 16 And
MBy >= (scaled_base_ line + 15) / 16 and
MBy < (scaled_base_ line + scaled_base_ height) / 16
Inter-layer motion prediction
A high layer macroblock can exploit inter-layer prediction using scaled base layer motion data, using either "BASE_LAYER_MODE" or "QPEL_REFINEMENT_MODE", as in case of dyadic spatial scalability. In case of using one of these two modes, the high layer macroblock is reconstructed with default motion data deduced from the base layer. These modes are only authorized for high layer macroblocks having corresponding base layer blocks, i.e. grey-colored area in figure 3 where bold line represents the upsampled base layer window 3.
- base-layer prediction macroblock construction
As in [JSVM1.0], these macrob lock modes indicate that motion/prediction information including macroblock partitioning are directly derived from the base layer. The process consists in constructing a prediction macroblock, MB pred, inheriting motion data from base layer. When using "BASE_LAYER_MODE" mode, the macroblock partitioning as well as the reference indices and motion vectors are those of the prediction macroblock MBjpred. "QPEL_REFINEMENT_MODE" is similar, but a quarter-sample motion vector refinement is achieved.
The process to derive MBjpred works in three steps:
- for each 4x4 block of MBjpred, inheritance of motion data from the base layer motion data
- partitioning choice for each 8x8 block of MBjpred
- mode choice for MBjpred
These 3 steps are described in the following sections.
- 4x4 block inheritance
For each 4x4 block b of the macroblock, the inheriting process consists in considering the center point (xc, yc) of the block, computing its base layer luma location (xB, yB), and inheriting motion data from the base layer block containing the pixel (xB, yB). Figure 4 represents the localisation of a high layer pixel in the base layer.
The detailed motion data inheritance process for b is the following:
Identification of base layer corresponding macroblock, 8x8 block and 4x4 block :
- Let (xθ,yθ) be the luma coordinates of the upper left pixel of b. The base luma location ( xB, yB ) of the pixel (1 ,1 ) of b in the base layer is derived as follows :
_ (xθ - scaled base column + 1) * BasePicWidthInSamplesL scaled base width
(yθ - scaled base line + 1) * BasePicHeightInSamplesL
Figure imgf000006_0001
scaled base height with
BasePicWidthlnSamplesL= 16*( BasePicWidthlnMbsMinus1+1 ) BasePicHeightlnSamplesL=16*( BasePicHeightlnMbsMinus1+1 ) the base corresponding macroblock is defined as the base layer macroblock containing pixel ( xB, yB ). Its coordinates ( mbxB, mbyB ) are given by the following equations:
Figure imgf000007_0001
- the base corresponding 8x8 block is defined as the base layer 8x8 block, belonging to the base corresponding macroblock, and containing pixel ( xB, yB ). Its coordinates ( bδxB, bδyB ) in the base corresponding macroblock are defined by the following equations: ibSxB = (xB - mbxB * 16) / 8 <
[bSyB = (yB - mbyB * 16) / 8
- the base corresponding 4x4 block is defined as the base layer 4x4 block containing pixel ( xB, yB ). Its coordinates ( b4xB, b4yB ) in the base corresponding macroblock are defined by the following equations:
Figure imgf000007_0002
Motion data inheritance :
- IF the base corresponding macroblock is intra, THEN 4x4 block is set as intra
- ELSE for each list listx (listx=0 or 1 ), the 4x4 block gets the reference index and motion vector from the base corresponding 4x4 block
- 8x8 block inheritance Once each 4x4 block has been treated, a merging process is applied to merge reference indices (1 per list) and motion vectors of the 8x8 block it belongs to. In the following, 4x4 blocks of a 8x8 block are identified as indicated in figure 5, blocks b1 and b2 for the top, b3 and b4 for the bottom, from the left to the right. For each 8x8 block B, the following process is applied: - IF the four 4x4 blocks have been classified as intra blocks, B is considered as intra 8x8 block.
- ELSE, reference indices and partitioning mode are chosen as follows:
Reference indices choice (for assigning same indices to each 4x4 block) o for each list Ix
■ IF no 4x4 block uses this list, no reference index and motion vector of this list are set to B
■ ELSE • reference index rB(lx) for B is computed as the minimum of the existing reference indices of the 4 4x4 blocks: rB 0x) = . min (rb (lx)) be{b1;b2,b3,b4}
• IF (rb1(lx) != rβ(/x)) o rb1 (Ix) = r B(IX) o IF (rb2(lx)==rB(lx)) mvb1(lx) = mvb2(lx) o ELSE IF (rω(/x)==rβ(/x)) mι/M(/x)=mι/ω(/x) o ELSE IF (rM(lx)==rB(lx)) mvb1(lx)=mvM(lx)
• IF (rb2(lx) != rB(lx)) o rb2(lx) = rB(lx) o IF (rb1(lx)==rB(lx)) mvb2(lx) = mvb1(lx) o ELSE IF (rM(lx)==rB(lx)) mvb2(lx)=mvM(lx) o ELSE IF (rω(/x)==rβ(/x)) mv^lx^mv^lx)
• IF (rω(/x) != rB(lx)) o rω(/x) = rB(lx) o IF (rM(lx)==rB(lx)) mvb3(lx) = mvM(lx) o ELSE IF (rb1(lx)==rB(lx)) mι/ω(/x)=mι/M(/x) o ELSE IF (rb2(lx)==rB(lx)) mι/ω(/x)=mι/62(/x)
• IF (rM(lx) != rB(lx)) o rM(lx) = rB(lx) o IF (rω(/x)==rβ(/x)) mvM(lx) = mv^lx) o ELSE IF (rb2(lx)==rB(lx)) mvM(lx)=mvb2(lx) o ELSE IF (rb1(lx)==rB(lx)) mvM(lx)=mvb1(lx)
Choice of partitioning mode o Two 4x4 blocks are considered as identical if their motion vectors are identical. The merging process is applied as follows:
■ IF bi is identical to b2 and b3 is identical to b4 THEN
• if bi is identical to b3 then BLK_8x8 is chosen • else BLK_8x4 is chosen
■ ELSE IF bi is identical to ύ3 and b2 is identical to b4 THEN BLK_4x8 is chosen
■ ELSE BLK_4x4 is chosen
- prediction macroblock mode choice
A final process is achieved to determine the MB pred mode. In the following, 8x8 blocks of the macroblock are identified as indicated in figure 6, blocks B1 and B2 for the top, B3 and B4 for the bottom, from the left to the right.
Two 8x8 blocks are considered as identical blocks if: - One or both of the two 8x8 blocks are classified as intra blocks or
- Partitioning mode of both blocks is BLK_8x8 and reference indices and motion vectors of listO and Iist1 of each 8x8 block, if they exist, are identical.
The mode choice is done using the following process:
- IF all 8x8 blocks are classified as intra blocks, THEN MB pred is classified as INTRA macroblock
- ELSE, MB pred is an INTER macroblock Remaining intra 8x8 blocks are enforced to be Inter 8x8 blocks o 8x8 blocks classified as intra are enforced to INTER blocks with 8x8 partitioning. Their reference indices and motion vectors are computed as follows. Let B,NTRA be such a 8x8 block. for each list listx
■ IF none of the other 8x8 block uses this list, no reference index and motion vector of this list is assigned tO BiNTRA
■ ELSE, the following steps are applied: • reference index r(listx) is computed as the minimum of the existing reference indices of the other 8x8 blocks: r (listx) = (listx))
Figure imgf000010_0001
• mean motion vector mvmean(listx) of the 8x8 blocks having the same reference index r(listx) is computed
• r(listx) is assigned to BINTRA and each 4x4 block of BiNTRA is enforced to have r{listx) and mvmean(listx) as reference index and motion vector. Mode choice o The choice of the partitioning mode for B is achieved. Two 8x8 blocks are considered as identical if their Partitioning mode is
8x8 and reference indices and motion vectors of listO and Iist1 of each 8x8 block, if they exist, are identical. The merging process is applied as follows:
■ if B1 is identical to B2 and B3 is identical to B4 then
• if B1 is identical to B3 then MODE_16x16 is chosen.
• else MODE_16x8 is chosen. ■ else if B1 is identical to B3 and B2 is identical to B4 then
MODE_8x16 is chosen.
■ else MODE 8x8 is chosen. - motion vectors scaling
A motion vector rescaling is finally applied to every existing motion vectors of the prediction macroblock MB pred. Motion data mv=(dχ,dy) is scaled in the mvs=(dsx,dSy) using the following equations:
|dsx = [dx .αhoriz] |dsy = [dy .αverticj
Using the semantic described further, the actual formulas are as follows :
2 * dx * scaled base width + sign[dx ] * BasePicWidthInSamplesL dSχ = 2 * BasePicWidthInSamplesL
2 *dy * scaled base height + sign[dy J* BasePicHeightInSamplesL sy 2 * BasePicHeightlnSamples L
Inter-layer texture prediction
Inter layer texture prediction is based on the same principles as inter layer motion prediction. It is only possible for macroblocks fully embedded in the scaled base layer window (grey-colored area in figure 3). For Intra texture prediction, the interpolation filter is applied across transform blocks boundaries. For residual texture prediction, this process only works inside transform blocks (4x4 or 8x8 depending on the transform).
The process at the decoder works as follows. Let MB be a high layer texture macroblock to be interpolated. Texture samples of MB are derived as follows:
- let ( xP , yP ) be the position of the upper left pixel of the macroblock in the high layer coordinates reference.
- a base layer prediction array is first derived as follows: o the corresponding quarter-pel position ( x4, y4 ) of ( xP , yP ) in the base layer is computed as: x4 = 4*x/>*BasePicWidthInSamplesL /scaled_base_width [y4 = 4 *yP* BasePicHeightlnSamples L /scaled_base_height the integer-pel position ( xB , yB ) is then derived as:
Figure imgf000012_0001
the quarter-pel phase is then derived as:
Figure imgf000012_0002
o the base layer prediction array corresponds to the samples contained in the area (xB-8, yB-8) and (xB+16, yB+16). The same filling process, as used in the dyadic case, is applied to fill samples areas corresponding to non existing or non available samples (for instance, in case of intra texture prediction, samples that do not belong to intra blocks). - the base layer prediction array is then upsampled. The upsampling is applied in two steps : first, texture is upsampled using the AVC half pixel 6-tap filter; then a bilinear interpolation is achieved to build the quarter pel samples, which results in a quarter-pel interpolation array. For intra texture, this interpolation crosses block boundaries. For residual texture, interpolation does not cross transform block boundaries.
- then, prediction sample pred[ x, y ] at each position ( x, y ), x=0..N- 1 ,y=0..N-1 , of the high layer block is computed as:
pred[ x, y ] = interp[ xl , yl ] with xl = px + (4 * x * BasePicWidthInSamplesL /scaled base width) [yl = py + (4 * y * BasePicHeightInSamplesL /scaled base height) interp[ xl , yl ] is the quarter-pel interpolated base layer sample at position ( xl, yl )
- inter-layer intra texture prediction
A given macroblock MB of current layer can exploit intra layer residual prediction only if co-located macroblocks of the base layer exist and are intra macroblocks.
For generating the intra prediction signal for high-pass macroblocks coded in I_BL mode, the corresponding 8x8 blocks of the base layer high-pass signal are directly de-blocked and interpolated, as in case of 'standard' dyadic spatial scalability. The same padding process is applied for deblocking. - inter-layer residual prediction
A given macroblock MB of current layer can exploit inter layer residual prediction only if co-located macroblocks of the base layer exist and are not intra macroblocks.
At the encoder, the upsampling process consists in upsampling each elementary transform block, without crossing the block boundaries. For instance, if a MB is coded into four 8x8 blocks, four upsampling processes will be applied on exactly 8x8 pixels as input.
The interpolation process is achieved in two steps : first, the base layer texture is upsampled using the AVC half pixel 6-tap filter; then a bilinear interpolation, is achieved to build the quarter pel samples. Interpolated high layer samples The nearest quarter pel position is chosen as the interpolated pixel.
Figure 7 shows how the 4x4 blocks are oversampled in the case of a 4/3 ratio. The number of pixels per block in the original picture is 4x4. After the 4/3 oversampling, this number becomes 6x6, 6x5, 5x6 or 5x5 depending on the location of the block within the picture. It can be observed that this number of pixels inside an oversampled 4x4 block follows a regular periodic pattern. In the 4/3 case, the pattern is : {6, 5, 5}
The phase of the top left pixel of each, expressed as the position of this pixel relatively to original grid is given by another periodic pattern. In the 4/3 case, the pattern is : {0, 1/2, 1/4}.

Claims

1 Process for scalable coding of images with a F1 format (1 ) corresponding to an enhanced layer and with a F2 format (2) having a lower resolution than F1 , corresponding to a base layer, comprising an inter layer residual prediction mode to code a current macroblock of the enhancement layer from a prediction macroblock of the base layer, a macroblock being made up of elementary blocks, the prediction macroblock being obtained through an upsampling of the base layer (3), characterised in that the upsampling is made at an elementary block level and in that the phase of the top left sample within an elementary block depends on the location of the elementary block within the picture.
2 Process according to claim 1 , characterized in that the size of an elementary block is 4 x 4 pixels.
3 Process according to claim 1 , characterized in that the filter used for the upsampling is a 6-tap filter.
4 Process according to claim 1 , characterized in that the upsampling is a 4/3 upsampling corresponding to 4 samp les for 3 original pixels.
5 Process according to claim 4, characterized in that the phase, expressed as the relative position of the pixel, for 3 successive elementary blocks, is 0, VT. and ΛA.
PCT/EP2006/061563 2005-04-13 2006-04-12 Process for scalable coding of images Ceased WO2006108863A2 (en)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
EP05075870 2005-04-13
EP05075870.5 2005-04-13

Publications (1)

Publication Number Publication Date
WO2006108863A2 true WO2006108863A2 (en) 2006-10-19

Family

ID=36808354

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/EP2006/061563 Ceased WO2006108863A2 (en) 2005-04-13 2006-04-12 Process for scalable coding of images

Country Status (1)

Country Link
WO (1) WO2006108863A2 (en)

Cited By (8)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
WO2007132077A1 (en) * 2006-05-16 2007-11-22 Arkema France Catalytic composition comprising catalytic activated carbon and carbon nanotubes, manufacturing process, electrode and super capacitator comprising the catalytic compound
FR2915048A1 (en) * 2007-04-16 2008-10-17 France Telecom Oversampling method for decoding scalable data stream in data transmission system, involves determining sub pixel position of given pixel from position of reference pixel and from offset representing distance between positions of pixels
US7742532B2 (en) 2006-11-17 2010-06-22 Lg Electronics Inc. Method and apparatus for applying de-blocking filter to a video signal
US8054885B2 (en) 2006-11-09 2011-11-08 Lg Electronics Inc. Method and apparatus for decoding/encoding a video signal
US8401085B2 (en) 2006-09-07 2013-03-19 Lg Electronics Inc. Method and apparatus for decoding/encoding of a video signal
WO2014161355A1 (en) * 2013-04-05 2014-10-09 Intel Corporation Techniques for inter-layer residual prediction
JP2016530774A (en) * 2013-07-15 2016-09-29 クゥアルコム・インコーポレイテッドＱｕａｌｃｏｍｍ Ｉｎｃｏｒｐｏｒａｔｅｄ Cross-layer parallel processing and offset delay parameters for video coding
US10093725B2 (en) 2010-08-19 2018-10-09 Zoetis Belgium S.A. Anti-NGF antibodies and their use

Cited By (14)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
FR2901156A1 (en) * 2006-05-16 2007-11-23 Arkema Sa Catalytic composition, useful to cover the electrode collectors and as electrode like electrochemical supercondenser, comprises polymeric binder and catalytic compound comprising carbon nanotubes and catalytic activated carbon
WO2007132077A1 (en) * 2006-05-16 2007-11-22 Arkema France Catalytic composition comprising catalytic activated carbon and carbon nanotubes, manufacturing process, electrode and super capacitator comprising the catalytic compound
US8428144B2 (en) 2006-09-07 2013-04-23 Lg Electronics Inc. Method and apparatus for decoding/encoding of a video signal
US8401085B2 (en) 2006-09-07 2013-03-19 Lg Electronics Inc. Method and apparatus for decoding/encoding of a video signal
US8054885B2 (en) 2006-11-09 2011-11-08 Lg Electronics Inc. Method and apparatus for decoding/encoding a video signal
US7742532B2 (en) 2006-11-17 2010-06-22 Lg Electronics Inc. Method and apparatus for applying de-blocking filter to a video signal
US8184698B2 (en) 2006-11-17 2012-05-22 Lg Electronics Inc. Method and apparatus for decoding/encoding a video signal using inter-layer prediction
US8229274B2 (en) 2006-11-17 2012-07-24 Lg Electronics Inc. Method and apparatus for decoding/encoding a video signal
US7742524B2 (en) 2006-11-17 2010-06-22 Lg Electronics Inc. Method and apparatus for decoding/encoding a video signal using inter-layer prediction
FR2915048A1 (en) * 2007-04-16 2008-10-17 France Telecom Oversampling method for decoding scalable data stream in data transmission system, involves determining sub pixel position of given pixel from position of reference pixel and from offset representing distance between positions of pixels
US10093725B2 (en) 2010-08-19 2018-10-09 Zoetis Belgium S.A. Anti-NGF antibodies and their use
WO2014161355A1 (en) * 2013-04-05 2014-10-09 Intel Corporation Techniques for inter-layer residual prediction
US10045041B2 (en) 2013-04-05 2018-08-07 Intel Corporation Techniques for inter-layer residual prediction
JP2016530774A (en) * 2013-07-15 2016-09-29 クゥアルコム・インコーポレイテッドＱｕａｌｃｏｍｍ Ｉｎｃｏｒｐｏｒａｔｅｄ Cross-layer parallel processing and offset delay parameters for video coding

Similar Documents

Publication Publication Date Title
CN101204092B (en) Method for deriving encoding information of high-resolution image from low-resolution image and encoding and decoding device implementing the method
JP5213456B2 (en) Method for deriving encoding information of high resolution picture from low resolution picture, and encoding and decoding apparatus for realizing the method
JP4751928B2 (en) Method and system for image level adaptation in extended spatial scalability
JP5112332B2 (en) Inter-layer motion prediction method
JP5011378B2 (en) Simplified inter-layer motion estimation for scalable video coding
CN101073265B (en) Scalable Video Coding Method
US20070140354A1 (en) Methods and Systems for Block-Based Residual Upsampling
KR20160108430A (en) Reference layer sample position derivation for scalable video coding
GB2505728A (en) Inter-layer Temporal Prediction in Scalable Video Coding
Kundu Light field compression using homography and 2D warping
CA3155627A1 (en) Encoder and decoder, encoding method and decoding method for reference picture resampling extensions
WO2006108863A2 (en) Process for scalable coding of images
EP1694074A1 (en) Process for scalable coding of images
EP2005760B1 (en) Method of predicting motion and texture data
Marquant et al. Extended spatial scalability for non dyadic video formats: from SDTV to HDTV
CN112534824A (en) Method and apparatus for video encoding
EP1958447A2 (en) Method of predicting motion and texture data
EP1848218A1 (en) Method for deriving motion data for high resolution pictures from motion data of low resolution pictures and coding and decoding devices implementing said method
EP1958449B1 (en) Method of predicting motion and texture data

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application
DPE1 Request for preliminary examination filed after expiration of 19th month from priority date (pct application filed from 20040101)
NENP Non-entry into the national phase in:

Ref country code: DE

WWW Wipo information: withdrawn in national office

Country of ref document: DE

NENP Non-entry into the national phase in:

Ref country code: RU

WWW Wipo information: withdrawn in national office

Country of ref document: RU

122 Ep: pct app. not ent. europ. phase

Ref document number: 06743290

Country of ref document: EP

Kind code of ref document: A2

WWW Wipo information: withdrawn in national office

Ref document number: 6743290

Country of ref document: EP