WO2001033827A2 - Procede et dispositif pour transformer un bloc d'image constitue de points - Google Patents

Procede et dispositif pour transformer un bloc d'image constitue de points Download PDF

Info

Publication number
WO2001033827A2
WO2001033827A2 PCT/DE2000/003752 DE0003752W WO0133827A2 WO 2001033827 A2 WO2001033827 A2 WO 2001033827A2 DE 0003752 W DE0003752 W DE 0003752W WO 0133827 A2 WO0133827 A2 WO 0133827A2
Authority
WO
WIPO (PCT)
Prior art keywords
image
coding
image block
block
transformed
Prior art date
Application number
PCT/DE2000/003752
Other languages
German (de)
English (en)
Inventor
Aktiengesellschaft Siemens
Original Assignee
Riegel, Thomas
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 Riegel, Thomas filed Critical Riegel, Thomas
Publication of WO2001033827A2 publication Critical patent/WO2001033827A2/fr

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
    • H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
    • H04N19/649—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding the transform being applied to non rectangular image segments

Definitions

  • the invention relates to an arrangement and a method for transforming an image block from image block points into a predetermined figure.
  • Such an arrangement and such a method are usually used in the encoding and decoding of a digitized image from image blocks.
  • an image template is segmented into image blocks corresponding to objects that occur in a scene and each of these objects is encoded separately.
  • FIG. 7 shows a camera 701 with which images are recorded.
  • the camera 701 can be any analog camera 701 that records images of a scene and either digitizes the images itself and transmits the digitized images to a first computer 702, which is coupled to it, or transmits the images analogously to the first computer 702.
  • the digitized images are processed or the analog images are converted into digitized images and the digitized images are processed.
  • the camera 701 can be a digital camera 701, with which directly digitized images are recorded and fed to the first computer 702 for further processing.
  • the first computer 702 can also be designed as an independent arrangement with which the method steps described below are carried out, for example as an independent computer card that is installed in a further computer.
  • the first computer 702 is generally to be understood as a unit that can carry out image signal processing in accordance with the method described below, for example a mobile terminal (mobile phone with a screen),
  • the first computer 702 has a processor unit 704 with which the method steps of image coding and image decoding described below are carried out.
  • the processor unit 704 is coupled via a bus 705 to a memory 706 in which image information is stored.
  • the methods described below can be implemented both in software and in hardware or also partly in software and partly in hardware.
  • the image decoding is carried out in the second computer 708.
  • the second computer 708 can have the same structure as the first computer 701.
  • the second computer 708 thus also has a processor 709 which is coupled to a memory 710 by a bus 711.
  • FIG. 6 shows an arrangement in the form of a basic circuit diagram for picture coding or picture decoding.
  • the arrangement shown can be used in the context of block-based and, as explained in more detail below, in the context of object-based image coding.
  • a digitized image 601 is divided into usually square image blocks 620 of size 8x8 pixels 602 or 16x16 pixels 602 and fed to the arrangement 603 for image coding.
  • Coding information is usually uniquely assigned to a pixel 602, for example brightness information (luminance values) and / or color information (chrominance values).
  • the digitized picture 601 is respectively with the picture elements 602 of the digital coded coding information associated with talized image 601 and transmitted.
  • inter-picture coding only one difference picture information of two successive digitized pictures 601 is coded and transmitted.
  • the difference information is only very small if movements of image objects in the temporally successive digitized images 601 are small. If the movements are large, there is a lot of difference information that is difficult to code. For this reason, as is known from [3], an "image-to-image" movement (motion estimation) is measured and compensated for before the difference information is determined (motion compensation).
  • a block-matching method is usually used for block-based image coding. It is based on the fact that a picture block to be coded is compared with reference picture blocks of the same size of a reference picture.
  • the criterion for a match quality between the block to be coded and a reference picture block in each case is usually the sum of the absolute differences in coding information which is assigned to each picture element. In this way, motion information for the image block, for example a motion vector, is determined, which motion information is transmitted with the difference information.
  • Two switch units 604 are provided for switching between the intra-picture coding and the inter-picture coding.
  • a subtract Ons unit 605 is provided, in which the difference in image information between two successive digitized images 601 is formed.
  • the image coding is controlled by an image coding control unit 606.
  • the image blocks 620 or difference image blocks to be coded are each fed to a transformation coding unit 607.
  • the transformation coding unit 607 applies a transformation coding, for example a discrete cosine transformation (DCT), to the coding information assigned to the pixels 602.
  • DCT discrete cosine transformation
  • any transformation coding for example a discrete sine transformation or a discrete Fourier transformation, can be used for image coding.
  • Spectral coefficients are formed by the transformation coding.
  • the spectral coefficients are quantized in a quantization unit 608 and fed to an image coding multiplexer 621, for example for channel coding and / or entropy coding.
  • the quantized spectral coefficients are inversely quantized in an inverse quantization unit 609 and subjected to an inverse transformation coding in an inverse transformation coding unit 610.
  • picture information of the respective temporally preceding picture is added in an adding unit 611.
  • the images reconstructed in this way are stored in a memory 612.
  • a unit for the motion compensation 613 is symbolically represented in the memory 612 for the sake of simplicity.
  • a loop filter (loop filter) 614 is provided, which is connected to the memory 612 and the subtraction unit 605.
  • the image coding multiplexer 621 is supplied with a mode index, with which it is indicated in each case whether intra-image coding or inter-image coding has been carried out.
  • the image coding multiplexer 621 is supplied with quantization indices 616 for the spectral coefficients.
  • a motion vector is assigned to an image block 620 and / or a macro block 623, which has four image blocks 620, for example, and is supplied to the image coding multiplexer 621.
  • information is provided for activating or deactivating the loop filter 614.
  • the transmitted information can be decoded in a second computer 619.
  • an image decoding unit 625 is provided in the second computer 619, which has, for example, the construction of a reconstruction loop of the arrangement shown in FIG.
  • a shape-adapted transformation coding is known, such as is used in particular in the context of object-based picture coding on edge picture blocks or picture blocks which contain only partially relevant coding information.
  • the edge image blocks coded using a shape-adapted transformation coding are characterized in that only those pixels are coded that are assigned to an object or have coding information relevant to the object.
  • the method described in [4] is a so-called shape-adjusted discrete cosine transformation ( ⁇ hape-Adaptive DCT, SA-DCT).
  • the transformation coefficients assigned to an image object are determined in such a way that pixels of an edge image block that do not belong to the image object are hidden.
  • a one-dimensional DCT the length of which corresponds to the number of remaining pixels in the respective column, is then first applied to the remaining pixels.
  • the resulting transformation coefficients are aligned horizontally and then subjected in a further one-dimensional DCT in the horizontal direction with the corresponding length.
  • the SA-DCT regulation known from [4] is based on a DCT-N transformation matrix with the following structure:
  • N denotes a size of the image vector to be transformed, in which the transformed pixels are contained.
  • DCT-N denotes a transformation matrix of size NxN.
  • indices are designated with p, k e [0, N-1].
  • each column of the image block to be transformed is made according to the regulation Cj * [DCT - N (p, k)] * Xj (2)
  • Various methods are used in computer graphics to display an object on a screen.
  • One method of representing an object is the so-called texture mapping.
  • Pixels which each contain brightness information (luminance values) and / or color information (chrominance values) of the object to be displayed, are mapped onto a surface of a three-dimensional model of the object to be displayed.
  • the three-dimensional model 301 of the object to be displayed which model 301 is shown in FIG. 3, consists of a spatial triangular lattice structure 301, the corner points 302 of the triangles 303 being present as points 304 in a Cartesian coordinate system 305.
  • the coordinates of the corner points are stored together with structure information, which structure information describes in each case an association of a corner point 302 with the respective, associated triangle 303.
  • each triangle 303 is assigned a so-called block-shaped structure map 306. net, which is made up of rectangular or block-shaped pixels 307. Brightness information (luminance values) and / or color information (chrominance values) are usually assigned to each pixel 307.
  • the triangle 303 is assigned the brightness or color information in such a way that an associated image point 307 of the associated structure map 306 is assigned to a corner point 302 and 308 of the triangle 303 and 309.
  • the position of a corner point 308 of the triangle 309 is determined by the
  • the coordinates (ui, vj_) 310 are usually standardized.
  • each corner point 302 of each triangle 303 of the three-dimensional model 301 is given a corresponding point 310, usually again a corner point, in the associated one
  • Structure card 306 assigned.
  • texture information is stored, which in each case describes an association of a corner point of a triangle in the structure map with the respective associated triangle.
  • all structure maps 401 are combined to form a digitized image 402, a so-called super structure map 402, in that the individual structure maps 401 are arranged in rows and columns. If necessary, the structure cards 401, which contain coding information relevant for the representation of the object, must be supplemented with structure cards 404, which contain no coding information relevant for the representation of the object. It is also known from [5] that such a superstructure map, as generated in the context of texture mapping, is encoded and decoded during image transmission.
  • the coding and / or decoding of a superstructure card is usually carried out using a block-oriented transformation in the intra-picture coding mode, as described above.
  • the described procedure for processing a digital image, in particular for the transformation of an image block, is not very efficient with regard to a lower data rate to be aimed at for transmission or a higher image quality.
  • the invention is therefore based on the problem of specifying a method and an arrangement for transforming an image block from image block points, with which a more efficient processing of a digitized image is possible.
  • Image block points the image block is transformed into a predetermined figure, the image block points being transformed using an invertible image.
  • a processor is provided which is set up in such a way that the image block can be transformed into a predetermined figure, the image block points being transformed using an invertible image.
  • An invertible image is to be understood as an image whose inverse image is unique in the mathematical sense (unambiguous).
  • the particular advantage of the invention is that when coding an image block transformed using an invertible image, no additional texture information has to be stored or coded.
  • the procedure according to the invention simplifies the coding of a digitized image compared to the known coding methods described above.
  • a maximum transmission capacity of a transmission channel can thus be used more effectively, i.e. a larger amount of user data can be transmitted.
  • the given figure describes a triangular shape.
  • the triangular shape preferably has a right angle.
  • the triangular shape using a first value and a second value.
  • the first value can describe a first side length of the triangular shape and the second value can describe a second side length of the triangular shape.
  • a corner point of the picture block is assigned to a corner point of the transformed picture block in such a way that the assignment is invertible or unambiguous.
  • coding information usually brightness information and / or color information, is assigned to an image block point.
  • the image block or the transformed image block is then encoded using the coding information.
  • the coding is preferably also carried out using the first, second and third values.
  • SA-DCT SA-DCT Transformation
  • a further simplification is obtained if a triangle adaptive discrete cosine transformation (TA-DCT) is used for coding and / or an inverse TA-DCT is used for decoding.
  • TA-DCT triangle adaptive discrete cosine transformation
  • Show it: 1 shows an arrangement for image coding and image decoding with a picture of an object by means of a camera and a representation of the object on a screen
  • Figure 2 Schematic representation of the procedure for image coding and image decoding with a picture of an object by means of a camera and a representation of the object on a screen
  • Figure 3 triangular lattice structure of the three-dimensional model with an associated structure map
  • Figure 5 Representation of a transformation of a structure map to a triangular, rectangular structure map
  • FIG 6 Sketch of an arrangement for block-based image coding or image decoding
  • Figure 7 Arrangement for image coding or image decoding with a camera, two computers and a transmission medium
  • FIG. 1 shows an arrangement for image coding and image decoding with a picture of an object by means of a camera and a representation of the object on a screen.
  • FIG. 1 shows a camera 101 with which images of an object 152 are recorded.
  • the camera 101 is an analog color camera, which records images of the object 152 and transmits the images in analog form to a first computer 102.
  • the analog images are converted into digitized images, wherein pixels of the digitized images contain color information of the object 152, and the digitized images are processed.
  • the object 152 is centered on a slide 153.
  • the relative position of the slide 153 with respect to the camera 101 is fixed.
  • By rotating the ob- Object carrier 153 can be moved around its center in such a way that, with the object 152 remaining at a constant distance from the camera 101, the viewing angle at which the camera 101 records the object 152 changes continuously.
  • the first computer 102 is designed as an independent arrangement in the form of an independent computer card, which is installed in the first computer 102, with which computer card the method steps described below are carried out.
  • the first computer 102 has a processor 104 with which the method steps of image coding described below are carried out.
  • the processor unit 104 is coupled via a bus 105 to a memory 106 in which image information is stored.
  • the method for image coding described below is implemented in software. It is stored in the memory 106 and is executed by the processor 104.
  • the image decoding is carried out in the second computer 108. Then, using the decoded image information of the object 152, a model of the object 152 is displayed on a screen 155 linked to the second computer 108.
  • the second computer 108 has the same structure as the first computer 101.
  • the second computer 108 also has a processor 109, which processor is coupled to a bus 110 with a memory 110.
  • the method described below for image decoding is implemented in software. It is stored in memory 110 and executed by processor 109.
  • FIG. 2 schematically shows the procedure for processing a digitized image in the context of coding and decoding with the recording of an object by means of a camera and a representation of the object on a screen.
  • This procedure for coding and decoding is implemented by the arrangement shown in FIG. 1 and the arrangement described above.
  • Step 1 recording the object (201)
  • images of the object 152 which is rotated in its position with respect to the camera 101 by means of the specimen slide 153, are recorded.
  • the images are transmitted in analog form to the first computer 102.
  • the camera 101 is calibrated, as described in [7], a spatial geometry of the arrangement and the recording parameters of the camera 101, for example the focal length of the camera 101, being determined.
  • the geometry data and the camera parameters are transmitted to the first computer 102.
  • the processing of the digitized images 103 takes place according to the method of automatic three-dimensional modeling using several images of an object, as described in [7].
  • a volume model 301 of the object 152 is determined using a method for determining a contour of an object in a digitized image, as described in [7], using the camera parameters and the digitized images 103.
  • the volume model 301 of the object 152 is described using a spatial triangular lattice structure 301, the corner points 302 of the triangles 303 being points 304 in a Cartesian coordinate system 305.
  • the volume model 301 is stored using the corner points 302 and structural information.
  • the corner points 302 of the lattice triangles 301 are numbered in each case.
  • the three numbers of the corner points 302 belonging to an ith lattice triangle 303 are each considered an ith number triplet (number of the first corner point / number of the second corner point / number of the third corner point)! (i: index for a grid triangle) saved.
  • a so-called structure map 306 or 502 is determined for each triangle 303 using the digitized images 103 and the color information contained in pixels of the digitized images 103.
  • the structure map 306 or 502 is constructed from block-shaped, square pixels 307 or 513.
  • One side length of the square image block has five pixels 307 and 513.
  • Each pixel 307 contains color information (chrominance values) of the object 152.
  • the color information is assigned to a grid triangle 303 in that an associated pixel 308 of the associated square structure map 306 is assigned to a corner point 302 of the grid triangle 303.
  • pixels of the square structure map 306 or 502 which pixels contain color information relevant for the representation of the object 152, are transformed into a new triangular structure map 503 from pixels.
  • the transformation which is shown in FIG. 5 and is explained below, is carried out using an invertible image.
  • the pixels 506 of the new triangular structure card 510 are arranged in such a way that they form a right-angled and isosceles triangle, one leg each having five pixels 506.
  • the transformation is carried out in such a way that the pixels which are corner pixels 504, 507, 508 of the triangle 505 in the square structure map 502 (i-th number triplet), with pixels, which are corner pixels 507, 511, 512 of the triangular structure map 510 , to match.
  • the corner point 508 mentioned first in the ith triplet of numbers is transformed into the corner point 507 of the triangular structure card 510, in which leg of the triangular structure card 510 enclose a right angle.
  • the corner point 509 mentioned second in the ith triplet of numbers is transformed to the corner point 512 of the triangular structure card 510, which lies on the right leg.
  • the corner point 504 mentioned third in the ith triplet of numbers is transformed to the corner point 511 of the triangular structure card 510, which lies on the left leg.
  • the pixels 501 which lie within the triangle spanned by the i-th triplet of numbers, are transformed such that ratios of distances a: b: c of the pixel 501 to the corner points 504, 508, 509 of the triangle 505 in the square structure map 502 correspond to the ratios of the distances a ⁇ : b ⁇ : c ⁇ of the pixel 506 to the corner pixels 507, 511, 512 of the triangular structure map 510.
  • pixels may have to be generated by linear extrapolation or a linear interpolation of values containing the color information, or pixels may have to be deleted.
  • a so-called triangle-adaptive discrete-cosine transformation (TA-DCT) is used for coding a triangular structure card.
  • This method for coding a digitized image is based on the method of a shape-adaptive discrete-cosine transformation (SA-DCT), as described in [4].
  • the transformation coefficients assigned to an image object are determined in such a way that pixels of an edge image block that do not belong to the image object are hidden.
  • a one-dimensional DCT the length of which corresponds to the number of remaining pixels in the respective column, is then first applied to the remaining pixels.
  • the resulting transformation coefficients are then subjected to a further one-dimensional DCT in the horizontal direction with a corresponding length.
  • the TA-DCT process is based on a DCT-N transformation matrix with the following structure:
  • N denotes a size of the image vector to be transformed, in which the transformed pixels are contained.
  • DCT-N denotes a transformation matrix of size NxN.
  • indices are designated with p, k e [0, N-1].
  • each column of the image block to be transformed is in accordance with the regulation
  • T-DCT triangle adaptive discrete cosine transformation
  • the image information coded using the TA-DCT (image information of the triangular structure cards) is transmitted to the second computer 108 via a transmission medium 107 together with data of the volume model of the object and the structure information of the volume model.
  • An image decoding is carried out after transmission.
  • the spectral coefficients cj are fed to an inverse TA-DCT.
  • the inverse TA-DCT as part of the image coding in the intra-
  • Image coding mode are pixels XJ from the spectral
  • - N denotes a size of the image vector to be transformed, in which the pixels to be transformed are contained
  • p, k indices are denoted by p, ke [0, Nl]; - () _1 denotes an inversion of a matrix.
  • the triangular structure maps are reconstructed.
  • the model of the object 152 is displayed on the screen 108.
  • Rectangular instead of triangular, ⁇ and isosceles structural map can be any triangular and rectangular structural map are used.
  • Such a triangular and rectangular structure map is described by a first value, which indicates the number of pixels of a first triangle side of the structure map, and a second value, which indicates the number of pixels of a second triangle side of the structure map.
  • the first value and the second value are saved and transferred for each structure map.
  • the triangular, rectangular structure maps can each be assigned a third value which specifies a number of color channels which are used for a color representation of a grid triangle of the object to be displayed, which triangle is associated with a structure map.
  • the third value is saved and transferred for all triangular, rectangular structure cards.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)

Abstract

L'invention concerne un procédé et un dispositif pour transformer un bloc d'image constitué de points en une figure prédéterminée, au moyen d'une représentation pouvant être inversée.
PCT/DE2000/003752 1999-10-29 2000-10-24 Procede et dispositif pour transformer un bloc d'image constitue de points WO2001033827A2 (fr)

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
DE19952196A DE19952196A1 (de) 1999-10-29 1999-10-29 Verfahren und Anordnung zur Transformation eines Bildblocks aus Bildblockpunkten
DE19952196.419991029 1999-10-29

Publications (1)

Publication Number Publication Date
WO2001033827A2 true WO2001033827A2 (fr) 2001-05-10

Family

ID=7927349

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/DE2000/003752 WO2001033827A2 (fr) 1999-10-29 2000-10-24 Procede et dispositif pour transformer un bloc d'image constitue de points

Country Status (2)

Country Link
DE (1) DE19952196A1 (fr)
WO (1) WO2001033827A2 (fr)

Also Published As

Publication number Publication date
DE19952196A1 (de) 2001-05-10

Similar Documents

Publication Publication Date Title
DE69831961T2 (de) Bildobjekterzeugungsverfahren für objektbasierte kodierungssysteme unter verwendung von masken und gerundeten mittelwerten
DE69824554T2 (de) Verfahren und anordnung zum erzeugen eines standbildes mit hoher auflösung
DE69723550T2 (de) Kodierung und dekodierung von grafischen symbolen
DE69634962T2 (de) Extrapolation von Pixelwerten eines in einem Block enthaltenen Videoobjektes
DE69730713T2 (de) Bildkoder und -dekoder
DE69728649T2 (de) Verfahren zur ergänzung eines digitalen bildes mit bildelementen
DE102013015821B4 (de) System und Verfahren zur Verbesserung der Videokodierung unter Verwendung von Inhaltsinformation
EP0956539B1 (fr) Procede et dispositif de codage et de decodage d'une image numerisee
DE602004001993T2 (de) Transformations basiertes restbewegungsrahmen kodierungsverfahren mit übervollständiger basis und zugehörige vorrichtung zur videokompression
DE69915843T2 (de) Teilbandkodierung/-dekodierung
EP0956703B1 (fr) Procede et dispositif de codage et de decodage d'une image numerisee
EP0985317B1 (fr) Procede de codage et de decodage d'une image numerisee
EP1279291B1 (fr) Procede et dispositif de mise en memoire et de traitement d'informations d'images successives dans le temps
EP0336510B1 (fr) Codeur predictif d'image immobile
EP1110407B1 (fr) Procede et dispositif pour le codage et le decodage d'une image numerisee faisant appel a un vecteur de deplacement total
DE19641157A1 (de) Verfahren zur Überprüfung der Konvergenz bei der fraktalen Bildcodierung
WO2001033827A2 (fr) Procede et dispositif pour transformer un bloc d'image constitue de points
WO2000025232A1 (fr) Procede et dispositif de codage, de decodage et de transmission d'une image numerisee
WO2000025520A1 (fr) Procede et dispositif pour le traitement d'une image numerisee
DE112019000210T5 (de) Würfelbasiertes Projektionsverfahren, welches unterschiedliche Abbildungsfunktionen auf unterschiedliche quadratische Projektionsflächen, unterschiedliche Achsen und/oder unterschiedliche Lagen einer Achse anwendet
EP0786186B1 (fr) Codage par transformation orthogonale bidimensionnelle pouvant s'adapter aux segments, pour un systeme de television numerique
EP1121809B1 (fr) Procede et dispositif pour le codage d'une image numerisee, procede et dispositif pour le decodage d'une image numerisee
EP0981909B1 (fr) Procede et dispositif de codage et de decodage d'une image numerisee
DE69005157T2 (de) Datencodierung.
DE3733038A1 (de) Verfahren und schaltungsanordnung zur bilddatenreduktion fuer digitale fernsehsignale

Legal Events

Date Code Title Description
AK Designated states

Kind code of ref document: A2

Designated state(s): CN JP KR US

AL Designated countries for regional patents

Kind code of ref document: A2

Designated state(s): AT BE CH CY DE DK ES FI FR GB GR IE IT LU MC NL PT SE

121 Ep: the epo has been informed by wipo that ep was designated in this application
WA Withdrawal of international application