WO2001003071A1 - Codage d'image 3d - Google Patents

Codage d'image 3d Download PDF

Info

Publication number
WO2001003071A1
WO2001003071A1 PCT/US2000/016156 US0016156W WO0103071A1 WO 2001003071 A1 WO2001003071 A1 WO 2001003071A1 US 0016156 W US0016156 W US 0016156W WO 0103071 A1 WO0103071 A1 WO 0103071A1
Authority
WO
WIPO (PCT)
Prior art keywords
geometry value
error difference
geometry
predicted
value
Prior art date
Application number
PCT/US2000/016156
Other languages
English (en)
Inventor
Francis M. L. Ng
Boon-Lock Yeo
Minerva Ming-Yee Yeung
Original Assignee
Intel Corporation
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 Intel Corporation filed Critical Intel Corporation
Priority to AU54841/00A priority Critical patent/AU5484100A/en
Publication of WO2001003071A1 publication Critical patent/WO2001003071A1/fr

Links

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
    • G06T9/00—Image coding
    • G06T9/001—Model-based coding, e.g. wire frame

Definitions

  • the present invention relates to the field of computer graphics.
  • the present invention discloses a method for efficiently encoding three-dimensional graphical models.
  • each triangular surface is represented by three vertices and three planar normals (one normal per vertex).
  • the vertex and normal coordinates of each triangle are stored as four-byte single precision floating-point values.
  • the memory bus data rate must be at least 4.3Gigabytes/second. Such a model ignores colors and texture coordinates that may consume even more memory bus bandwidth.
  • a method of encoding a three-dimensional graphics model is disclosed. The method first compares a geometry value from the three-dimensional graphics model with a predicted geometry value. The method then quantizes an error difference between said geometry value with a predicted geometry value to produce a quantized error difference.
  • Figure 1A illustrates a block diagram of a geometry encoder for the 3D model coding portion of the MPEG4 standard.
  • Figure IB illustrates a block diagram of a geometry decoder for the 3D model coding portion of the MPEG4 standard.
  • Figure 2A illustrates a block diagram of a geometry encoder built according to the teachings of the present invention.
  • Figure 2B illustrates a block diagram of a geometry decoder built according to the teachings of the present invention.
  • Figure 3A illustrates a wire frame rendering of a triangular mesh model of a power drill.
  • Figure 3B illustrates a graph of the bits/triangle is plotted against the corresponding maximum square quantization error for two different geometry encodings of the power drill of Figure 3A.
  • Figure 3C illustrates a graph of the bits/triangle is plotted against the corresponding mean square quantization error for two different geometry encodings of the power drill of Figure 3A.
  • Figure 4A illustrates a wire frame rendering of a triangular mesh model of a horse.
  • Figure 4B illustrates a graph of the bits /triangle is plotted against the corresponding maximum square quantization error for two different geometry encodings of the horse of Figure 4A.
  • Figure 4C illustrates a graph of the bits/triangle is plotted against the corresponding mean square quantization error for two different geometry encodings of the horse of Figure 4A.
  • Complex three-dimensional graphics requires large detailed three-dimensional object models.
  • Large three-dimensional object models put great demands on computer resources.
  • large three-dimensional object models require large amounts of permanent storage space for storage, large amounts of main memory for fast access while rendering, and a large memory bus bandwidth to move the three-dimensional object models to the processor for processing.
  • many three-dimensional graphics systems have adopted the approach of compressing the three-dimensional object models.
  • An ideal compression system for three-dimensional object models would provide a very good compression ratio, nmodel object distortion, and a simple implementation.
  • a group known as the Motion Picture Expert Group has been defining an image compression system known as "MPEG4" that compresses images by defining objects in the image with three-dimensional object models.
  • the compressed images can be reconstructed by rendering the scenes using the three-dimensional object models.
  • the MPEG4 standard needs to adopt a standard 3D model coding technique commonly referred to as 3DMC.
  • 3DMC 3D model coding
  • the 3D model coding technique proposed by Taubin and Rossignac (hereinafter referred to as 3DMC/TR) consists of two main components: a connectivity encoding component and geometry encoding component.
  • the connectivity encoding component attempts to reduce the redundancy inherent in many representations of polyhedral or triangular 3D meshes. For example, one straightforward but inefficient technique to represent a triangular mesh is to represent each triangle in the mesh by direct reference to the triangle's three vertices.
  • the 3DMC/TR system first decomposes a mesh into spanning trees of triangles and vertices. These trees are encoded separately in a lossless manner.
  • the geometry encoding component of 3DMC/TR performs lossy compression to reduce the size of geometric data.
  • the geometry encoding component compresses vertex coordinate, vertex normal, color, and texture coordinate data.
  • the present invention focuses on techniques to improve the geometry encoding of three-dimensional object model data. Thus, the remainder of this document will concentrate on geometry encoding.
  • Figure 1A illustrates a block diagram of the 3DMC/TR geometry encoding method.
  • the 3DMC/TR geometry encoding method first normalizes the vertex coordinate geometry into a unit cube (not shown) to produce a series of unit vertex coordinates U n .
  • a uniform scalar quantizer Q' quantizes the floating-point vertex coordinates U n into fixed length integers i n .
  • the uniform scalar quantizer Q' truncates each normalized value U multiplied by 2 L to produce an integer i, where L is the number of quantization levels for each coordinate.
  • the quantization can be defined by the formula ⁇ 2 J.
  • the 3DMC/TR geometry encoding method compares the quantized vertex coordinate i n with other nearby quantized coordinates. Since proximity in the vertex spanning tree often implies geometric proximity of the corresponding vertices, the recently quantized coordinates are compared with the just quantized coordinate. Specifically, a predictor P in Figure 1A produces linear combination of neighboring vertices a n from the vertex spanning tree to predict the current vertex position. The method then encodes the error difference e n between the predicted value (a n ) and the actual quantized vertex position . When vertex coordinates are quantized in a fixed-point representation scheme, these error difference eaus corrective vectors on average have a smaller magnitude than absolute positions.
  • error difference e n corrective vectors may be encoded in a lossless manner with fewer bits.
  • the error difference e n corrective terms are further compressed with the entropy coder 145 using standard lossless compression techniques to produce a final stream of compressed error difference e n .
  • FIG. IB illustrates a block diagram of a 3DMC/TR geometry decoder.
  • an entropy decoder 155 decompresses the compressed error difference e n to produce uncompressed error signal e n .
  • the predictor P then adds error signal e n to the prediction value a. to produce the quantized value .
  • the quantized value i Vietnamese is used by the Predictor P to produce future prediction values a n .
  • ⁇ n is dequantized to create the reconstructed geometry value U n .
  • One major drawback of the 3DMC/TR geometry encoding technique is that the direct quantization of input vertex coordinates produces a relatively high level of distortion.
  • the distortion of the encoded 3D models introduced by the 3DMC/TR geometry encoder may produce less than acceptable final images.
  • the present invention introduces an improved geometry encoding technique that provides better results than the 3DMC/TR geometry encoder.
  • the geometry encoding method of the present invention quantizes the error difference of input vertex coordinates and the associated prediction values instead of directly quantizing the input vertex coordinates. The quantized difference is later added back to the same prediction values to form the reconstructed values.
  • This technique is commonly known as differential pulse code modulation (DPCM) and is widely used in speech and audio coding.
  • DPCM differential pulse code modulation
  • Figure 2A illustrates a block diagram of a geometry encoder built according to the teachings of the present invention.
  • a predicted value U n created by predictor P from the previously reproduced values is subtracted from the input value U n to form an error signal e n .
  • a quantizer Q r then quantizes the error signal e n to produce a quantized error signal e n .
  • An entropy encoder compresses the quantized error signal e n to produce a stream of compressed error signals i,.
  • An adder creates a reconstructed value U n by adding the quantized error signal e n to the predicted value U n .
  • the predictor P uses one or more reconstructed values U n to form the next prediction value U n .
  • the predictor P may be the same type of predictor P that is used in the 3DMC/TR geometry coding system .
  • This type of DPCM encoding is based on one assumption. If the prediction value U n is reasonably accurate and a good approximation of the actual coordinate value U Titan, then the magnitude of the error signal e n is small and in most cases smaller than the actual coordinate value U n . In this case, the variance of error signal e n will be less than the variance of coordinate value U n . Therefore, quantizing the reduced dynamic range of the error signal e n with the same number of quantization levels produces better results than directly quantizing the actual coordinate value U n . Thus, the effective resolution of the quantizer is increased without increasing the number of quantization levels.
  • the same distortion can be achieved by reducing the number of quantization levels and keeping the distortion the same as would be used for direct quantization of U n .
  • the better "effective" resolution of the quantizer will provide a more accurate quantized error signal e n .
  • the present invention produces a more accurate reconstructed value U n , which in turn should give a more accurate prediction value U n .
  • the quantizer should be selected to provide a minimum average distortion for a fixed bit rate.
  • the quantizer is followed immediately by entropy coding.
  • the quantizer should be designed to operate with high resolution so that the number of quantization levels N is large, the probability of overload should be very small, and the step size is much smaller than the root mean square (rms) signal value.
  • rms root mean square
  • the dynamic range of the uniform quantizer must still be selected. If the dynamic range of the quantizer is chosen to match the dynamic range of the input values too closely, low distortion is achieved at the expense of high bit rate according to equation (1). Furthermore, it quite difficult to estimate the dynamic range of the input values without a second pass. To avoid the risk of overloading the quantizer (one of the underlying assumptions for the optimality of the uniform quantizer followed by entropy coding), the dynamic range of the quantizer must be carefully selected. Specifically, the quantizer dynamic range should be larger than the input data but not too large such that the reduction in distortion is too small for the scheme to be of any practical interest.
  • the present invention proposes using "rounding" quantizer Q r .
  • the rounding quantizer Q' rounds off floating point values to the nearest fixed-length integers instead of truncating the value.
  • the quantization scheme Q' used in 3DMC/TR geometry encoder proposed to the MPEG4 committee truncates the floating point values. This truncation produces more quantization error than rounding off floating point values.
  • Figure 2B illustrates a block diagram of the corresponding geometry decoder of the present invention.
  • the entropy encoded (compressed) quantized prediction error is first decompressed to produce an uncompressed quantized prediction error ,.
  • the quantized prediction error i_ is dequantized to form a prediction error value e n .
  • An adder then adds the prediction error eome to a prediction geometry value U n from predictor P to generate a final reconstructed geometry value U titanium.
  • the geometry decoder of the present invention is actually identical to a portion of the corresponding geometry encoder except for the inverse quantizer, by comparing the dotted-line region in Figure 2B and the dotted-line region in Figure 2A.
  • the 3DMC/TR geometry decoder can be converted to a DPCM geometry decoder with just a few modifications. Specifically, one only needs to change the order of the quantizer block and the predictor block to change a 3DMC/TR geometry decoder into a DPCM decoder.
  • the dynamic range of the quantizer used in the encoder may be inserted into the bitstream such that the decoder is aware of the encoder's dynamic range.
  • a fixed dynamic range may be preset to preclude any change in the 3DMC syntax.
  • the first column of Table 1 lists the names of the three-dimensional object models that were tested.
  • Figures 3A and 4A illustrate a wire-frame rendering of the Power Drill and Horse three-dimensional models, respectively.
  • the second column of Table 1 lists the number of triangles for the different object models.
  • the third and forth columns of Table 1 list the percentage of bit rate reduction when the DPCM geometry encoder of the present invention is used instead of the standard geometry encoder of MPEG4 3DMC.
  • the third and forth columns of Table 1 show that the DPCM geometry encoder with the rounding quantizer Q r consistently outperforms the proposed 3DMC/TR geometry encoder in bit rate for a given maximum square quantization error and mean square quantization error.
  • the percentage reductions in bit rate with DPCM and rounding quantizer Q r can range from 2% to 20% using the criterion of maximum square quantization error and from 4% to 20% using the criterion of mean square quantization error.
  • the absolute reductions in bit rate with DPCM and rounding quantizer Q' is almost constant in the regions of interest, so the higher percentage reduction corresponds to the lower bit rate region while the lower percentage reduction corresponds to the higher bit rate region.
  • Figures 3B and 4B illustrate graphs of the bits/triangle is plotted against the corresponding maximum square quantization error for the different geometry encodings of the power drill of Figure 3A.
  • Figures 3C and 4C illustrate graphs of the bits/triangle is plotted against the corresponding mean square quantization error for the different geometry encodings of the power drill of Figure 3A.
  • the DPCM geometry encoder of the present invention consistently outperforms the 3DMC/TR system.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Physics & Mathematics (AREA)
  • General Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)

Abstract

La présente invention concerne un procédé destiné à coder avec efficacité la géométrie d'objets de modèles graphiques tridimensionnels. Ce procédé consiste à comparer une valeur de géométrie à partir du modèle graphique tridimensionnel avec une valeur prédite de géométrie créée à partir de prédictions de géométrie préalables, puis à quantifier une différence d'erreur entre la valeur réelle de géométrie et la valeur prédite de géométrie afin d'obtenir une différence d'erreur quantifiée, et enfin à coder par entropie la différence d'erreur quantifiée.
PCT/US2000/016156 1999-06-30 2000-06-12 Codage d'image 3d WO2001003071A1 (fr)

Priority Applications (1)

Application Number Priority Date Filing Date Title
AU54841/00A AU5484100A (en) 1999-06-30 2000-06-12 3d image coding

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
US34361399A 1999-06-30 1999-06-30
US09/343,613 1999-06-30

Publications (1)

Publication Number Publication Date
WO2001003071A1 true WO2001003071A1 (fr) 2001-01-11

Family

ID=23346822

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/US2000/016156 WO2001003071A1 (fr) 1999-06-30 2000-06-12 Codage d'image 3d

Country Status (3)

Country Link
AU (1) AU5484100A (fr)
TW (1) TW463135B (fr)
WO (1) WO2001003071A1 (fr)

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2839439A4 (fr) * 2012-04-18 2016-06-01 Thomson Licensing Procédé et dispositif de correction de sommet pour composants tridimensionnels (3d) ayant subi une rotation
US9866840B2 (en) 2013-01-10 2018-01-09 Thomson Licensing Method and apparatus for vertex error correction

Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0861003A2 (fr) * 1997-02-25 1998-08-26 SHARP Corporation Dispositif de contrÔle de la quantité de données et codeur l'utilisant
JPH10304362A (ja) * 1997-05-01 1998-11-13 Canon Inc 画像信号処理装置及び方法

Patent Citations (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP0861003A2 (fr) * 1997-02-25 1998-08-26 SHARP Corporation Dispositif de contrÔle de la quantité de données et codeur l'utilisant
JPH10304362A (ja) * 1997-05-01 1998-11-13 Canon Inc 画像信号処理装置及び方法

Non-Patent Citations (3)

* Cited by examiner, † Cited by third party
Title
AHN J -H ET AL: "Geometry compression of 3D models using adaptive quantization for prediction errors", PICTURE CODING SYMPOSIUM '99, PROCEEDINGS OF PICTURE CODING SYMPOSIUM, PORTLAND, OR, USA, 21-23 APRIL 1999, 1999, Corvallis, OR, USA, Oregon State Univ, USA, pages 193 - 197, XP000934376 *
LI J ET AL: "PROGRESSIVE CODING OF 3-D GRAPHIC MODELS", PROCEEDINGS OF THE IEEE,US,IEEE. NEW YORK, vol. 86, no. 6, June 1998 (1998-06-01), pages 1052 - 1063, XP000669753, ISSN: 0018-9219 *
PATENT ABSTRACTS OF JAPAN vol. 1999, no. 02 26 February 1999 (1999-02-26) *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
EP2839439A4 (fr) * 2012-04-18 2016-06-01 Thomson Licensing Procédé et dispositif de correction de sommet pour composants tridimensionnels (3d) ayant subi une rotation
US9621924B2 (en) 2012-04-18 2017-04-11 Thomson Licensing Vextex correction method and apparatus for rotated three-dimensional (3D) components
US9866840B2 (en) 2013-01-10 2018-01-09 Thomson Licensing Method and apparatus for vertex error correction

Also Published As

Publication number Publication date
TW463135B (en) 2001-11-11
AU5484100A (en) 2001-01-22

Similar Documents

Publication Publication Date Title
US8411975B2 (en) Method and apparatus for encoding and decoding key value data of coordinate interpolator
US7430330B2 (en) Method and apparatus for optimizing a JPEG image using regionally variable compression levels
AU684013B2 (en) Compact source coding tables for encoder/decoder system
EP1697900B1 (fr) Traitement d'image multi-mode
JP2005520360A (ja) 損失の無い方法でデジタル画像を符号化するための装置および方法
US6330369B1 (en) Method and apparatus for limiting data rate and image quality loss in lossy compression of sequences of digital images
JP2007116725A (ja) データアレイの予測符号化方法およびシステム
AU748947B2 (en) Method and apparatus for inverse quantization of MPEG-4 video
US20030133512A1 (en) Spatially transcoding a video stream
JP2008510349A (ja) グラフィックとビデオとの混合ソースを圧縮するシステムおよび方法
US20020131645A1 (en) Method and apparatus for optimizing a JPEG image using regionally variable compression levels
US7809204B2 (en) Method and apparatus for encoding and decoding key value data of coordinate interpolator
Kim et al. Fractal coding of video sequence using circular prediction mapping and noncontractive interframe mapping
KR100438856B1 (ko) 능동적인 테이블 생성에의한 양자화/역양자화 방법 및 장치
Al-Khafaji et al. Grey-Level image compression using 1-d polynomial and hybrid encoding technique
WO2001003071A1 (fr) Codage d'image 3d
US7099523B2 (en) Method and system for scaling a signal sample rate
Teng et al. A new quadtree predictive image coder
CN100566419C (zh) 以无损的方式编码数字图像数据的设备和方法
Sim et al. An efficient 3D mesh compression technique based on triangle fan structure
Zhao et al. A JPEG codec adaptive to region importance
KR20020069983A (ko) 삼차원 객체의 형태 변형 정보에 대한 부호화 방법 및 그장치
Tsern et al. Image coding using pyramid vector quantization of subband coefficients
Ng et al. On DPCM and its variation applied to geometry compression
Bethel et al. Optimal quantisation of the discrete cosine transform for image compression

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A1

Designated state(s): AE AG AL AM AT AU AZ BA BB BG BR BY CA CH CN CR CU CZ DE DK DM DZ EE ES FI GB GD GE GH GM HR HU ID IL IN IS JP KE KG KP KR KZ LC LK LR LS LT LU LV MA MD MG MK MN MW MX MZ NO NZ PL PT RO RU SD SE SG SI SK SL TJ TM TR TT TZ UA UG US UZ VN YU ZA ZW

AL Designated countries for regional patents

Kind code of ref document: A1

Designated state(s): GH GM KE LS MW MZ SD SL SZ TZ UG ZW AM AZ BY KG KZ MD RU TJ TM AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE BF BJ CF CG CI CM GA GN GW ML MR NE SN TD TG

121 Ep: the epo has been informed by wipo that ep was designated in this application
DFPE Request for preliminary examination filed prior to expiration of 19th month from priority date (pct application filed before 20040101)
REG Reference to national code

Ref country code: DE

Ref legal event code: 8642

122 Ep: pct application non-entry in european phase
NENP Non-entry into the national phase

Ref country code: JP