WO2010146769A1 - Image encoding device, image encoding method, image decoding device, image decoding method, and image display device - Google Patents

Image encoding device, image encoding method, image decoding device, image decoding method, and image display device Download PDF

Info

Publication number
WO2010146769A1
WO2010146769A1 PCT/JP2010/003341 JP2010003341W WO2010146769A1 WO 2010146769 A1 WO2010146769 A1 WO 2010146769A1 JP 2010003341 W JP2010003341 W JP 2010003341W WO 2010146769 A1 WO2010146769 A1 WO 2010146769A1
Authority
WO
WIPO (PCT)
Prior art keywords
image
quantization
representative point
gradation
encoding
Prior art date
Application number
PCT/JP2010/003341
Other languages
French (fr)
Japanese (ja)
Inventor
村上智一
Original Assignee
株式会社日立製作所
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 株式会社日立製作所 filed Critical 株式会社日立製作所
Priority to JP2011519503A priority Critical patent/JP5342645B2/en
Publication of WO2010146769A1 publication Critical patent/WO2010146769A1/en

Links

Images

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
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
    • H04N19/154—Measured or subjectively estimated visual quality after decoding, e.g. measurement of distortion
    • 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
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/124—Quantisation
    • 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
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
    • H04N19/136—Incoming video signal characteristics or properties
    • H04N19/14—Coding unit complexity, e.g. amount of activity or edge presence estimation
    • 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
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
    • H04N19/17—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object
    • H04N19/176—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object the region being a block, e.g. a macroblock

Definitions

  • the present invention relates to an image encoding device, an image encoding method, an image decoding device, an image decoding method, and an image display device that efficiently encode and decode an image.
  • a coding system such as the MPEG (Moving Picture Experts Group) system has been established, and the MPEG-1 standard, MPEG-2 standard, MPEG-4 standard, etc. This is a standard encoding method.
  • H.264 / AVC Advanced Video Video Coding
  • Patent Document 1 uses a multilevel image which is a grayscale image such as a photograph by binarization processing by an error diffusion method and uses a binary image by an arithmetic encoding method.
  • the encoding technique is disclosed.
  • Patent Document 2 as a method of efficiently quantizing and digitizing an image signal expressed in a gray scale or a three-dimensional color space, the image signal is converted into a luminance / color difference separation type uniform color space, and a luminance axis is obtained.
  • a technique is disclosed in which the converted image signal is uniformly quantized for the color difference plane, and the converted image signal is quantized based on a triangular lattice for the color difference plane.
  • Patent Document 1 The technique described in Patent Document 1 is intended for a grayscale image and cannot be applied to a color image as it is. If each color component of a color image is processed independently using this method, the color balance is lost and color noise is generated, which significantly deteriorates subjective image quality.
  • the present invention has been made in view of the above problems, and an object of the present invention is to provide an image encoding / decoding technique that enables efficient encoding without greatly degrading subjective image quality.
  • an image encoding device includes a representative point determination unit that determines a plurality of representative points representing pixel values of an input image in the luminance / color difference space, and an input image in the luminance / color difference space.
  • a gradation vector quantization unit that performs vector quantization of the gradation of the input image using a representative point that is the shortest distance from the processing target pixel, information on the pixel value of the representative point, and quantization information of the input image A representative point to be encoded / quantized information encoding unit;
  • the gradation vector quantization unit weights the quantization error when the peripheral pixels of the processing target pixel are vector-quantized according to the positional relationship with the processing target pixel, and the total value of the quantization error is diffused. In this manner, the pixel value of the processing target pixel is corrected, and a representative point having the shortest distance with respect to the corrected pixel value is selected.
  • the image decoding apparatus of the present invention inputs an encoded stream obtained by encoding an image by vector quantization of the gradation of the image using a representative point in the luminance / chrominance space. From the representative point / quantization information decoding unit that decodes the pixel value information of the representative point and the quantization information of the image, the pixel value information of the representative point and the quantization information of the image, A gradation vector restoring unit that restores a gradation vector in the luminance / color difference space of the pixel, and an image quality enhancement filter that performs a filtering process on the restored gradation vector.
  • FIG. 1 is a configuration diagram (Example 1) showing an example of an image encoding device according to the present invention.
  • FIG. The figure which shows calculation of the error vector by the gradation vector quantization part 15.
  • FIG. The figure which shows the quantization of the pixel value by the gradation vector quantization part 15.
  • FIG. 3 is a flowchart showing a flow of an image encoding method in Embodiment 1.
  • FIG. The figure which shows the image processed by the image quality improvement filter 25.
  • FIG. 10 is a flowchart illustrating a flow of an image decoding method according to the second embodiment.
  • FIG. 1 is a block diagram showing an embodiment of an image encoding device according to the present invention.
  • the image encoding device 1 includes an image input unit 11 that inputs an image, an area determination unit 12 that determines the characteristics of an area in the image, an image encoding unit 13 that encodes an image using a general method, A representative point determination unit 14 for determining a representative point of a pixel in the luminance / chrominance space, a gradation vector quantization unit 15 for three-dimensionally quantizing each pixel using the representative point in the luminance / chrominance space, and a representative A representative point / quantization information encoding unit 16 that encodes the pixel value of the point and the quantized pixel information, and a code integration unit 17 that integrates the encoded stream are provided.
  • the image input unit 11 holds the input image and passes it to the region determination unit 12.
  • the region determination unit 12 analyzes the image feature for each processing target region of the image and determines the processing method.
  • the area of the image may be a block of a specific size, or may be an aggregate of these blocks or an area determined by preprocessing such as area division.
  • an image has a characteristic for each region, and compression efficiency can be further increased by switching an appropriate encoding method for each region.
  • the vector quantization of the gradation is performed three-dimensionally on the luminance and color difference information.
  • efficient encoding is performed without significantly degrading the subjective image quality.
  • the image feature determination method for example, dispersion values of color components are used.
  • the variance value is an average value of square errors with respect to the average value of each element.
  • region A a region where this variance value is smaller than a predetermined value
  • encoding is performed by gradation vector quantization in a luminance / chrominance space.
  • a general encoding method is used.
  • efficient encoding can be performed without significantly degrading the subjective image quality by performing gradation vector quantization.
  • the determination of the region is also performed when the region where the representative point is switched is separated during the gradation vector quantization described later.
  • the image encoding unit 13 performs encoding on the region B determined to be generally encoded by the above region determination. For example, H.I. An existing international standard encoding method such as H.264 / AVC is used.
  • the encoded image data is transferred to the code integration unit 17.
  • the representative point determination unit 14 determines a representative point for color information in the region for the region A to be encoded using the gradation vector quantization.
  • the number of representative points may be determined in advance, or may be determined according to color variations in the region. Hereinafter, an example using four representative points will be described.
  • As a representative point determination method pixel values are plotted in a color space, and a statistical method such as voting or clustering is used. Details of the representative point determination method will be described later.
  • Information on the determined representative point is sent to the gradation vector quantization unit 15. In an area where the distribution of pixel values changes in the image, more appropriate vector quantization is possible by switching the representative point to be selected.
  • the gradation vector quantization unit 15 performs vector quantization of the gradation of each pixel using a representative point in the luminance / color difference space. That is, the position of each processing target pixel is approximated using any of the four representative points, and information (quantization information) indicating which representative point is used is generated. The gradation vector quantization method will be described later.
  • the quantization information of each pixel is sent to the representative point / quantization information encoding unit 16.
  • the representative point / quantization information encoding unit 16 encodes the luminance / color difference information of the representative point and the quantization information for each pixel. For example, when there are four representative points, four points of Y, U, and V information are encoded. Further, if the representative points are four points, the quantization information of each pixel can be expressed by 2 bits in order to distinguish them. This is scanned with respect to the coding region, and entropy coding is performed using Huffman coding, arithmetic coding, or the like. Alternatively, encoding by vector quantization may be used in which a pattern is mapped to a code in units of about 4 ⁇ 4 size, and a pattern with a low appearance frequency is converted into another similar pattern and encoded. When the representative point is switched depending on the region, information on the luminance / color difference of the representative point is encoded in each region unit. The encoded representative point information and quantization information are sent to the code integration unit 17.
  • the vector coding of the gradation using the representative point is performed, thereby performing the conventional coding. Encoding can be performed more efficiently than the encoding method.
  • FIGS. 2A to 2C are diagrams for explaining in detail the operations of the representative point determination unit 14 and the gradation vector quantization unit 15 in the present embodiment.
  • the representative point determination unit 14 determines the gradation vector quantization unit 15 based on the four representative points.
  • FIG. 2A is a diagram showing determination of representative points.
  • the value of each pixel of the input image is plotted in a luminance / color difference space with Y, U, and V as axes, and each pixel value is represented by a vector P (indicated by a black circle).
  • vectors P0, P1, P2, and P3 representing these pixel values are determined (indicated by white circles).
  • a general clustering method or the like can be used as a representative point determination method. For example, an appropriate initial value may be given to P0, P1, P2, and P3, and the center of gravity may be calculated and adjusted each time a new value is plotted.
  • FIG. 2B is a diagram showing calculation of an error vector.
  • the vector of the pixel to be processed (marked with *) is P *, and the vectors of its surrounding pixels (a, b, c, d) are Pa, Pb, Pc, Pd.
  • representative points have already been selected for the peripheral pixels and are approximated by P0, P2, P2, and P3, respectively.
  • FIG. 2C is a diagram illustrating quantization of pixel values.
  • a representative point having the shortest distance is selected as the representative point of the processing target pixel.
  • the position of the processing target pixel is corrected so that the error vector of the surrounding pixels is diffused (error diffusion method).
  • a total value of error vectors (quantization errors) associated with vector quantization of peripheral pixels is obtained.
  • the range of peripheral pixels and the weighting coefficient may be the same as those in a general error diffusion method.
  • the error diffusion method is taken as an example.
  • a gradation quantization method used for a two-dimensional image such as a matrix dither method can also be used.
  • FIG. 3 is a flowchart showing the flow of the image encoding method in the present embodiment.
  • step S101 an original image is input, and in step S102, the screen is divided into a plurality of processing areas (for example, block areas).
  • step S103 the process moves to the processing target area.
  • step S104 the region determination unit 12 obtains the dispersion value of the color component of the target region and determines the processing method. That is, when the variance value is smaller than the predetermined value, the process proceeds to step S105, and gradation vector quantization in the luminance / color difference space is performed. If the variance value is larger than the predetermined value, the process proceeds to step S108, and the image encoding unit 13 performs image encoding in the area by a general method.
  • step S105 the representative point determination unit 14 determines a representative point from the luminance / color difference information of the pixels in the region.
  • step S106 the gradation vector quantization unit 15 performs vector quantization of the gradation of each pixel using the representative point.
  • step S107 the representative point / quantization information encoding unit 16 encodes the representative point and quantization information of each pixel.
  • step S109 the data encoded in S107 or S108 is stored in a memory.
  • step S110 it is determined whether or not the processing for all areas has been completed. If there is an unprocessed area, the process returns to step S103 to move to the next process target area and continue the encoding process.
  • step S111 the encoded data stored in the memory by the code integration unit 17 is integrated and output as an encoded stream. At this time, a flag indicating which encoding method is used for each region is added to the encoded stream.
  • FIG. 4 is a block diagram showing an embodiment of an image decoding apparatus according to the present invention.
  • the image decoding device 2 receives a coded stream and separates a code, a code separation unit 21, an image decoding unit 22 that decodes an image by a general method, and a representative point / quantization information decoding unit
  • the representative point / quantization information decoding unit 23 a gradation vector restoration unit 24 for restoring the gradation vector of each pixel in the luminance / chrominance space, a high image quality filter 25 for performing filtering on the restored image, and an image
  • An image composition unit 26 for composition and an image output unit 27 for outputting an image are provided.
  • the code separation unit 21 receives an encoded stream, and is encoded by a gradation vector quantization in a region (region B) encoded by a general method or a luminance / chrominance space by a given flag or the like. It is determined whether it is an area (area A).
  • the image decoding unit 22 decodes the encoded data determined to be the region B by a general method, and generates a decoded image. For this, an existing image decoding method (for example, H.264 / AVC) corresponding to the image encoding unit 13 of the image encoding device 1 is used.
  • the representative point / quantization information decoding unit 23 with respect to the encoded data determined to be the region A, information on luminance / color difference of the representative point and information indicating which representative point is used for each pixel (quantization information) Is decrypted. For this, a decoding scheme corresponding to the representative point / quantization information encoding unit 16 of the image encoding device 1 is used.
  • the gradation vector restoration unit 24 restores the gradation vector in the luminance / color difference space of each pixel using the luminance / color difference information of the representative point and the quantization information of each pixel.
  • the image quality enhancement filter 25 improves the subjective image quality by performing filter processing on the restored gradation vector (pixel value). This process may be performed independently for each color component, or may be performed collectively.
  • the simplest method is to apply a low pass filter to each color component.
  • the value of each pixel is quantized by the representative point, but the image quality can be improved by performing spatial interpolation by low-pass filter processing with surrounding pixels.
  • the image synthesis unit 26 synthesizes the image decoded by the image decoding unit 22 and the image output by the image quality enhancement filter 25, and the image output unit 24 outputs the synthesized image.
  • the image decoding apparatus 2 decodes an encoded stream efficiently encoded by vector quantization of gradations using representative points, and restores a high-quality decoded image. Is possible.
  • 5A to 5C are diagrams illustrating examples of images output from the gradation vector restoration unit 24 and the image quality improving filter 25 in the present embodiment.
  • FIG. 5A shows an original image to be decoded.
  • FIG. 5B is an image restored using the representative points in the gradation vector restoration unit 24. At this stage, since it is composed only of the pixel values of the representative points, it appears that when the pixels are enlarged, discrete values are arranged at random.
  • FIG. 5C is an image processed by the image quality enhancement filter 25.
  • filter processing with peripheral pixels is performed using a low-pass filter or an adaptive anisotropic diffusion filter in consideration of edges, and a high-quality image can be obtained as shown in the figure.
  • the improvement effect by the filter processing here is particularly effective when combined with the gradation vector quantization in the luminance / color difference space. That is, since representative points in the luminance / color difference space are selected and quantized three-dimensionally, the generation of color noise in the decoded image can be suppressed.
  • FIG. 6 is a flowchart showing the flow of the image decoding method in the present embodiment.
  • step S201 an encoded stream is input, and in step S202, the process moves to a processing target area on the screen.
  • step S203 the code separation unit 21 determines the encoding method of the target area based on the flag added to the stream.
  • step S204 the representative point / quantization information decoding unit 23 decodes the representative point and the quantization information of each pixel from the encoded stream.
  • step S205 the gradation vector restoration unit 24 restores the gradation vector of each pixel using the representative point in the luminance / color difference space.
  • step S206 the image restored by the image quality enhancement filter 25 is filtered.
  • step S208 the images generated in steps S206 and S207 are combined by the image combining unit 26.
  • step S209 it is determined whether or not the processing for all areas has been completed. If there is an unprocessed area, the process returns to step S202, moves to the next process target area, and continues the decoding process.
  • step S210 the image synthesized by the image output unit 27 is output.
  • FIG. 7 is a block diagram showing an embodiment of the image display device according to the present invention.
  • the image display device 3 inputs an encoded stream encoded by gradation vector quantization, decodes this, and displays an image on the display 33.
  • the configuration includes a representative point / quantization information decoding unit 31 that decodes representative point / quantization information, and a gradation vector restoration unit 32 that restores a gradation vector in the luminance / chrominance space. This is the same as the representative point / quantization information decoding unit 23 and the gradation vector restoration unit 24 in the second embodiment (FIG. 4). However, in this embodiment, the image quality enhancement filter 25 in FIG. 4 is omitted.
  • the image is decoded as in the second embodiment, but the filtering process is not performed on the decoded image.
  • the human vision will have a filtering effect on the decoded image. It is. That is, an image composed of only representative points as shown in FIG. 5B looks like an image after filtering as shown in FIG. 5C. This is particularly effective in the case of electronic paper devices that have a very small number of coloring functions and outdoor large display devices. In addition, it is possible to display an image with high image quality.
  • the image encoding device, the image decoding device, and the method thereof according to the present invention are widely applied to video processing devices such as an image recording device, a player, a mobile phone, a mobile terminal, a digital camera TV, a projector, various displays, and a game machine. be able to.

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)
  • Compression Of Band Width Or Redundancy In Fax (AREA)

Abstract

An image encoding device (1) is provided with: a representative point determining unit (14) which determines the representative points of pixel values of an input image in a luminance/color difference space; a gradation vector quantization unit (15) which uses the representative points closest to the pixels, of the input image, to be processed and performs vector quantization of the gradations of the input image; and an representative point/quantization information encoding unit (16) which encodes the information of the pixel values of the representative points and the quantization information of the input image. The gradation vector quantization unit (15) adds weight to the quantization error of the peripherals pixels around the pixels to be processed, corrects the pixel values of the pixels to be processed so that the quantization error diffuses, and selects the representative points with the closest distance to the corrected pixel values. Due to this, an image encoding/decoding technique capable of efficient encoding without deteriorating the subjective image quality can be provided.

Description

画像符号化装置、画像符号化方法、画像復号化装置、画像復号化方法、及び画像表示装置Image encoding device, image encoding method, image decoding device, image decoding method, and image display device
　本発明は、画像を効率的に符号化・復号化する画像符号化装置、画像符号化方法、画像復号化装置、画像復号化方法、及び画像表示装置に関する。 The present invention relates to an image encoding device, an image encoding method, an image decoding device, an image decoding method, and an image display device that efficiently encode and decode an image.
　画像、音声情報をデジタルデータ化して記録、伝送する手法として、ＭＰＥＧ（Moving Picture Experts Group）方式等の符号化方式が策定され、ＭＰＥＧ－１規格、ＭＰＥＧ－２規格、ＭＰＥＧ－４規格等として国際標準の符号化方式となっている。さらに圧縮率を向上させる方式として、Ｈ．２６４／ＡＶＣ（Advanced Video Coding）規格等が定められている。 As a technique for recording and transmitting image and audio information as digital data, a coding system such as the MPEG (Moving Picture Experts Group) system has been established, and the MPEG-1 standard, MPEG-2 standard, MPEG-4 standard, etc. This is a standard encoding method. As a method for further improving the compression ratio, H. The H.264 / AVC (Advanced Video Video Coding) standard is defined.
　一方、多値画像を符号化する方法として、特許文献１には、写真等の濃淡画像である多値画像を誤差拡散法によって２値化処理し、２値化画像を算術符号化方法を用いて符号化する技術が開示されている。
  また特許文献２には、グレースケールや３次元色空間で表現された画像信号を効率的に量子化してデジタル化する方法として、画像信号を輝度・色差分離型均等色空間へ変換し、輝度軸については変換した画像信号を均等量子化し、色差平面については変換した画像信号を三角格子に基づく量子化を行う技術が開示されている。 On the other hand, as a method of encoding a multilevel image, Patent Document 1 uses a multilevel image which is a grayscale image such as a photograph by binarization processing by an error diffusion method and uses a binary image by an arithmetic encoding method. The encoding technique is disclosed.
In Patent Document 2, as a method of efficiently quantizing and digitizing an image signal expressed in a gray scale or a three-dimensional color space, the image signal is converted into a luminance / color difference separation type uniform color space, and a luminance axis is obtained. A technique is disclosed in which the converted image signal is uniformly quantized for the color difference plane, and the converted image signal is quantized based on a triangular lattice for the color difference plane.
特開平５－２１９３８１号公報JP-A-5-219381 特開２００３－２８９５５１号公報JP 2003-289551 A
　特許文献１に記載の技術は、濃淡画像（グレースケール画像）を対象としており、このままカラー画像に適用することはできない。仮にこの手法を用いてカラー画像の各色成分を独立に処理した場合、色バランスが崩れて色ノイズが発生し、主観画質を大きく劣化させてしまう。 The technique described in Patent Document 1 is intended for a grayscale image and cannot be applied to a color image as it is. If each color component of a color image is processed independently using this method, the color balance is lost and color noise is generated, which significantly deteriorates subjective image quality.
　また特許文献２に記載の技術では、画像信号を輝度軸と色差平面の３次元色空間として取り扱うので、色ノイズの発生を抑えることが期待できる。しかしながら、この手法では処理対象の画像の特徴に関係なく一定の量子化レベルが設定されているので、必ずしも画像に適応した高い符号化効率が得られるとは限らない。 In the technique described in Patent Document 2, since the image signal is handled as a three-dimensional color space with a luminance axis and a color difference plane, generation of color noise can be expected. However, in this method, since a constant quantization level is set regardless of the characteristics of the image to be processed, high encoding efficiency adapted to the image is not always obtained.
　本発明は、上記課題に鑑みて為されたものであり、その目的は主観画質を大きく劣化させることなく、効率的に符号化が可能な画像符号化、復号化技術を提供することにある。 The present invention has been made in view of the above problems, and an object of the present invention is to provide an image encoding / decoding technique that enables efficient encoding without greatly degrading subjective image quality.
　上記目的を達成するために、本発明の画像符号化装置は、輝度・色差空間における入力画像の画素値を代表する複数の代表点を決定する代表点決定部と、輝度・色差空間における入力画像の処理対象画素と最短距離になる代表点を用いて、入力画像の階調のベクトル量子化を行う階調ベクトル量子化部と、代表点の画素値の情報と入力画像の量子化情報とを符号化する代表点・量子化情報符号化部とを備える。ここに階調ベクトル量子化部は、処理対象画素の周辺画素をベクトル量子化した際の量子化誤差に処理対象画素との位置関係に応じた重み付けを行い、量子化誤差の合計値が拡散するように処理対象画素の画素値を補正し、補正した画素値に対して最短距離になる代表点を選択する。 In order to achieve the above object, an image encoding device according to the present invention includes a representative point determination unit that determines a plurality of representative points representing pixel values of an input image in the luminance / color difference space, and an input image in the luminance / color difference space. A gradation vector quantization unit that performs vector quantization of the gradation of the input image using a representative point that is the shortest distance from the processing target pixel, information on the pixel value of the representative point, and quantization information of the input image A representative point to be encoded / quantized information encoding unit; Here, the gradation vector quantization unit weights the quantization error when the peripheral pixels of the processing target pixel are vector-quantized according to the positional relationship with the processing target pixel, and the total value of the quantization error is diffused. In this manner, the pixel value of the processing target pixel is corrected, and a representative point having the shortest distance with respect to the corrected pixel value is selected.
　また本発明の画像復号化装置は、輝度・色差空間における代表点を用いて画像の階調をベクトル量子化することで画像を符号化した符号化ストリームが入力するものであって、符号化ストリームから、代表点の画素値の情報と画像の量子化情報とを復号化する代表点・量子化情報復号化部と、代表点の画素値の情報と画像の量子化情報とを用いて、各画素の輝度・色差空間における階調ベクトルを復元する階調ベクトル復元部と、復元された階調ベクトルに対してフィルタ処理を行う高画質化フィルタと、を備える。 Also, the image decoding apparatus of the present invention inputs an encoded stream obtained by encoding an image by vector quantization of the gradation of the image using a representative point in the luminance / chrominance space. From the representative point / quantization information decoding unit that decodes the pixel value information of the representative point and the quantization information of the image, the pixel value information of the representative point and the quantization information of the image, A gradation vector restoring unit that restores a gradation vector in the luminance / color difference space of the pixel, and an image quality enhancement filter that performs a filtering process on the restored gradation vector.
　本発明によれば、主観画質を大きく劣化させることなく、効率的に符号化が可能な画像符号化、復号化技術を提供することができる。 According to the present invention, it is possible to provide an image encoding / decoding technique that enables efficient encoding without greatly degrading the subjective image quality.
本発明による画像符号化装置の一実施例を示す構成図（実施例１）。1 is a configuration diagram (Example 1) showing an example of an image encoding device according to the present invention. 代表点決定部１４による代表点の決定を示す図。The figure which shows the determination of the representative point by the representative point determination part 14. FIG. 階調ベクトル量子化部１５による誤差ベクトルの計算を示す図。The figure which shows calculation of the error vector by the gradation vector quantization part 15. FIG. 階調ベクトル量子化部１５による画素値の量子化を示す図。The figure which shows the quantization of the pixel value by the gradation vector quantization part 15. FIG. 実施例１における画像符号化方法の流れを示すフローチャート。3 is a flowchart showing a flow of an image encoding method in Embodiment 1. 本発明による画像復号化装置の一実施例を示す構成図（実施例２）。The block diagram which shows one Example of the image decoding apparatus by this invention (Example 2). 復号化対象の原画像を示す図。The figure which shows the original image of decoding object. 階調ベクトル復元部２４において代表点を用いて復元された画像を示す図。The figure which shows the image decompress | restored using the representative point in the gradation vector decompression | restoration part 24. FIG. 高画質化フィルタ２５により処理した画像を示す図。The figure which shows the image processed by the image quality improvement filter 25. FIG. 実施例２における画像復号化方法の流れを示すフローチャート。10 is a flowchart illustrating a flow of an image decoding method according to the second embodiment. 本発明による画像表示装置の一実施例を示す構成図（実施例３）。The block diagram which shows one Example of the image display apparatus by this invention (Example 3).
　以下、本発明の実施例を、図面を参照して説明する。 Embodiments of the present invention will be described below with reference to the drawings.
　図１は、本発明による画像符号化装置の一実施例を示す構成図である。画像符号化装置１は、画像を入力する画像入力部１１と、画像内の領域の特徴を判定する領域判定部１２と、一般的な方式で画像の符号化を行う画像符号化部１３と、輝度・色差空間における画素の代表点の決定する代表点決定部１４と、輝度・色差空間において代表点を用いて各画素を３次元的に量子化を行う階調ベクトル量子化部１５と、代表点の画素値と量子化された画素情報を符号化する代表点・量子化情報符号化部１６と、符号化ストリームを統合する符号統合部１７を備える。 FIG. 1 is a block diagram showing an embodiment of an image encoding device according to the present invention. The image encoding device 1 includes an image input unit 11 that inputs an image, an area determination unit 12 that determines the characteristics of an area in the image, an image encoding unit 13 that encodes an image using a general method, A representative point determination unit 14 for determining a representative point of a pixel in the luminance / chrominance space, a gradation vector quantization unit 15 for three-dimensionally quantizing each pixel using the representative point in the luminance / chrominance space, and a representative A representative point / quantization information encoding unit 16 that encodes the pixel value of the point and the quantized pixel information, and a code integration unit 17 that integrates the encoded stream are provided.
　以下では、色空間を輝度・色差空間として座標軸としてＹＵＶ（ＹＣｂＣｒ）を用いる場合を例に挙げて説明するが、空間としてはこれ以外に、ＲＧＢ、ＨＳＶ、Ｌ＊ａ＊ｂ＊、ＸＹＺ等の空間を用いることができる。 In the following, the case where YUV (YCbCr) is used as a coordinate axis with the color space as the luminance / color difference space will be described as an example. Space can be used.
　各部の動作を説明する。画像入力部１１は入力された画像を保持し、領域判定部１２に渡す。領域判定部１２は、画像の各処理対象領域について画像特徴を解析して処理方法の判定を行う。画像の領域は特定サイズのブロックとしてもよいし、これらブロックの集合体、あるいは領域分割等の前処理によって判別された領域としてもよい。 Explain the operation of each part. The image input unit 11 holds the input image and passes it to the region determination unit 12. The region determination unit 12 analyzes the image feature for each processing target region of the image and determines the processing method. The area of the image may be a block of a specific size, or may be an aggregate of these blocks or an area determined by preprocessing such as area division.
　一般に画像には領域毎に特徴があり、領域毎に適切な符号化方式を切り替えることにより圧縮効率をより高めることができる。特に、色成分の変化が少なく輝度成分の変化の大きいところでは、輝度及び色差情報について３次元的に階調のベクトル量子化を行う。これにより本実施例では、主観画質を大きく劣化させることなく効率的な符号化を行うものである。 Generally, an image has a characteristic for each region, and compression efficiency can be further increased by switching an appropriate encoding method for each region. In particular, where the change in the color component is small and the change in the luminance component is large, the vector quantization of the gradation is performed three-dimensionally on the luminance and color difference information. Thus, in this embodiment, efficient encoding is performed without significantly degrading the subjective image quality.
　画像特徴の判定方法としては、例えば色成分の分散値を用いる。分散値とは、各要素の平均値に対する二乗誤差の平均値である。この分散値が所定値より小さい領域では（これを領域Ａとする）、輝度・色差空間における階調ベクトル量子化による符号化を行い、分散値が所定値以上の領域であれば（領域Ｂとする）、一般的な符号化方式を用いる。これにより、色成分の変化が少ない領域Ａでは、階調ベクトル量子化を行うことで主観画質をそれほど劣化させずに効率的な符号化を可能とする。なお、領域の判定は、後述の階調ベクトル量子化の際に、代表点の切り替えを行う領域を分離する場合にも行われる。 As the image feature determination method, for example, dispersion values of color components are used. The variance value is an average value of square errors with respect to the average value of each element. In a region where this variance value is smaller than a predetermined value (this region is referred to as region A), encoding is performed by gradation vector quantization in a luminance / chrominance space. A general encoding method is used. As a result, in the area A where the change in the color component is small, efficient encoding can be performed without significantly degrading the subjective image quality by performing gradation vector quantization. The determination of the region is also performed when the region where the representative point is switched is separated during the gradation vector quantization described later.
　画像符号化部１３は、上記の領域判定によって一般的な符号化を行うと判定された領域Ｂに対して符号化を行う。これには、例えばＨ．２６４／ＡＶＣ等の既存の国際標準符号化方式を用いる。符号化された画像データは符号統合部１７に渡される。 The image encoding unit 13 performs encoding on the region B determined to be generally encoded by the above region determination. For example, H.I. An existing international standard encoding method such as H.264 / AVC is used. The encoded image data is transferred to the code integration unit 17.
　代表点決定部１４は、階調ベクトル量子化を用いて符号化する領域Ａに対して、領域内の色情報に対する代表点を決定する。代表点の数は予め決めておいてもよいし、領域内の色のばらつきに応じて決定してもよい。以下では、４点の代表点を用いる例で説明する。代表点の決定方法には、色空間内に画素値をプロットし、投票やクラスタリング等の統計的手法を用いる。代表点の決定方法の詳細については後述する。決定した代表点の情報は、階調ベクトル量子化部１５に送られる。なお、画像内部において画素値の分布が変化する領域では、選択する代表点を切り替えることでより適切なベクトル量子化が可能となる。 The representative point determination unit 14 determines a representative point for color information in the region for the region A to be encoded using the gradation vector quantization. The number of representative points may be determined in advance, or may be determined according to color variations in the region. Hereinafter, an example using four representative points will be described. As a representative point determination method, pixel values are plotted in a color space, and a statistical method such as voting or clustering is used. Details of the representative point determination method will be described later. Information on the determined representative point is sent to the gradation vector quantization unit 15. In an area where the distribution of pixel values changes in the image, more appropriate vector quantization is possible by switching the representative point to be selected.
　階調ベクトル量子化部１５は、輝度・色差空間における代表点を用いて各画素の階調のベクトル量子化を行う。すなわち、各処理対象画素の位置を４点の代表点のいずれかを用いて近似し、どの代表点を用いたかを示す情報（量子化情報）を生成する。階調のベクトル量子化方法については後述する。各画素の量子化情報は、代表点・量子化情報符号化部１６に送る。 The gradation vector quantization unit 15 performs vector quantization of the gradation of each pixel using a representative point in the luminance / color difference space. That is, the position of each processing target pixel is approximated using any of the four representative points, and information (quantization information) indicating which representative point is used is generated. The gradation vector quantization method will be described later. The quantization information of each pixel is sent to the representative point / quantization information encoding unit 16.
　代表点・量子化情報符号化部１６は、代表点の輝度・色差の情報と、各画素に対する量子化情報を符号化する。例えば代表点が４点だった場合、４点のＹ，Ｕ，Ｖの情報を符号化する。また各画素の量子化情報は、代表点が４点であれば、これらを区別するために２ビットで表現できる。これを符号化領域に対してスキャンし、ハフマン符号化、算術符号化等を用いてエントロピー符号化する。あるいは、４×４サイズ程度のブロック単位でパターンを符号にマッピングし、出現頻度の少ないパターンは似ている別のパターンに変換して符号化するベクトル量子化による符号化を用いてもよい。なお、代表点が領域によって切り替わる場合には、各領域単位で代表点の輝度・色差の情報を符号化する。符号化された代表点情報と量子化情報は、符号統合部１７に送られる。 The representative point / quantization information encoding unit 16 encodes the luminance / color difference information of the representative point and the quantization information for each pixel. For example, when there are four representative points, four points of Y, U, and V information are encoded. Further, if the representative points are four points, the quantization information of each pixel can be expressed by 2 bits in order to distinguish them. This is scanned with respect to the coding region, and entropy coding is performed using Huffman coding, arithmetic coding, or the like. Alternatively, encoding by vector quantization may be used in which a pattern is mapped to a code in units of about 4 × 4 size, and a pattern with a low appearance frequency is converted into another similar pattern and encoded. When the representative point is switched depending on the region, information on the luminance / color difference of the representative point is encoded in each region unit. The encoded representative point information and quantization information are sent to the code integration unit 17.
　符号統合部１７は、画像符号化部１３からの符号化データと、代表点・量子化情報符号化部１６からの符号化情報を統合し、符号化ストリームとして出力する。その際符号化ストリームには、対象領域ごとにいずれの符号化方式を用いたかを示すフラグを付与する。例えば、階調ベクトル量子化による場合はフラグ＝「１」とし、一般の符号化方式による場合はフラグ＝「０」とする。 The code integration unit 17 integrates the encoded data from the image encoding unit 13 and the encoded information from the representative point / quantization information encoding unit 16 and outputs the result as an encoded stream. At this time, a flag indicating which encoding method is used for each target region is assigned to the encoded stream. For example, when gradation vector quantization is used, flag = “1”, and when using a general encoding method, flag = “0”.
　以上のように本実施例の画像符号化装置１では、色成分の変化が少なく輝度成分の変化の大きい画像領域では、代表点を用いた階調のベクトル量子化を行うことで、従来の符号化方式に比較し効率的な符号化が可能となる。 As described above, in the image coding apparatus 1 according to the present embodiment, in the image region where the change of the color component is small and the change of the luminance component is large, the vector coding of the gradation using the representative point is performed, thereby performing the conventional coding. Encoding can be performed more efficiently than the encoding method.
　図２Ａ～図２Ｃは、本実施例における代表点決定部１４と階調ベクトル量子化部１５の動作を詳細に説明する図である。以下では、Ｙ，Ｕ，Ｖの輝度・色差空間において４点の代表点を決定し、これを基に誤差拡散法を利用して階調のベクトル量子化を行う例について説明する。 2A to 2C are diagrams for explaining in detail the operations of the representative point determination unit 14 and the gradation vector quantization unit 15 in the present embodiment. Hereinafter, an example will be described in which four representative points are determined in the Y, U, and V luminance / color difference spaces, and gradation vector quantization is performed using the error diffusion method based on the four representative points.
　図２Ａは代表点の決定を示す図である。まず、入力画像の各画素の値をＹ，Ｕ，Ｖを軸とする輝度・色差空間内にプロットし、各画素値をベクトルＰで表す（黒丸で示す）。次にこれらの画素値を代表するベクトルＰ０，Ｐ１，Ｐ２，Ｐ３を決定する（白丸で示す）。代表点の決定方法には、一般的なクラスタリング手法等を用いることができる。例えば、Ｐ０，Ｐ１，Ｐ２，Ｐ３に対して適当な初期値を与え、新しい値をプロットする毎に重心を計算して調整すればよい。単純な方法としては、図に示すようにＰ０の値を（Ｙ，Ｕ，Ｖ）＝（２５５，１２８，１２８）に固定し、Ｐ１，Ｐ２，Ｐ３についてはＹ＝０として、Ｕ，Ｖの平面内において各画素値によって投票を行い、Ｕ，Ｖの値を決定する。これによって代表点４点を決定する。 FIG. 2A is a diagram showing determination of representative points. First, the value of each pixel of the input image is plotted in a luminance / color difference space with Y, U, and V as axes, and each pixel value is represented by a vector P (indicated by a black circle). Next, vectors P0, P1, P2, and P3 representing these pixel values are determined (indicated by white circles). A general clustering method or the like can be used as a representative point determination method. For example, an appropriate initial value may be given to P0, P1, P2, and P3, and the center of gravity may be calculated and adjusted each time a new value is plotted. As a simple method, as shown in the figure, the value of P0 is fixed to (Y, U, V) = (255, 128, 128), and P = 0, P2, and P3 are set to Y = 0. Voting is performed by each pixel value in the plane, and the values of U and V are determined. As a result, four representative points are determined.
　図２Ｂは誤差ベクトルの計算を示す図である。処理対象画素（＊印）のベクトルをＰ＊、その周辺画素（ａ，ｂ，ｃ，ｄ）のベクトルをＰａ，Ｐｂ，Ｐｃ，Ｐｄとする。ここで周辺画素については既に代表点が選択されていて、それぞれＰ０，Ｐ２，Ｐ２，Ｐ３で近似されている。このとき各周辺画素には、選択された代表点との間で誤差ベクトルＶａ＝Ｐ０－Ｐａ，Ｖｂ＝Ｐ２－Ｐｂ，Ｖｃ＝Ｐ２－Ｐｃ，Ｖｄ＝Ｐ３－Ｐｄが生じている。 FIG. 2B is a diagram showing calculation of an error vector. The vector of the pixel to be processed (marked with *) is P *, and the vectors of its surrounding pixels (a, b, c, d) are Pa, Pb, Pc, Pd. Here, representative points have already been selected for the peripheral pixels and are approximated by P0, P2, P2, and P3, respectively. At this time, error vectors Va = P0−Pa, Vb = P2−Pb, Vc = P2−Pc, and Vd = P3−Pd are generated in each peripheral pixel with the selected representative point.
　図２Ｃは画素値の量子化を示す図である。処理対象画素の代表点には両者の距離が最短となる代表点を選択するが、まず、処理対象画素の位置について、周辺画素の誤差ベクトルが拡散されるように補正する（誤差拡散法）。始めに、周辺画素のベクトル量子化に伴う誤差ベクトル（量子化誤差）の合計値を求める。その際、対象画素＊との位置関係に応じた重み付けを行い、例えばその係数をａ：ｂ：ｃ：ｄ＝１：３：１：３とする。その結果、誤差ベクトルの合計値Ｖｔｏｔは、
      Ｖｔｏｔ＝Ｖａ＋３Ｖｂ＋Ｖｃ＋３Ｖｄ
となる。なお、周辺画素の範囲や重み付け係数は一般の誤差拡散法と同様でよい。 FIG. 2C is a diagram illustrating quantization of pixel values. A representative point having the shortest distance is selected as the representative point of the processing target pixel. First, the position of the processing target pixel is corrected so that the error vector of the surrounding pixels is diffused (error diffusion method). First, a total value of error vectors (quantization errors) associated with vector quantization of peripheral pixels is obtained. At that time, weighting is performed according to the positional relationship with the target pixel *, and for example, the coefficient is set to a: b: c: d = 1: 3: 1: 3. As a result, the total error vector Vtot is
Vtot = Va + 3Vb + Vc + 3Vd
It becomes. The range of peripheral pixels and the weighting coefficient may be the same as those in a general error diffusion method.
　次に、対象画素の画素値Ｐ＊を周辺画素の誤差ベクトルＶｔｏｔで補正する。補正値をＰｑとすると、
      Ｐｑ＝Ｐ＊＋Ｖｔｏｔ
となる。そして、Ｐｑと代表点であるＰ０，Ｐ１，Ｐ２，Ｐ３の輝度・色差空間内の距離を計算し、最短距離となる代表点を選択する。図の例では、Ｐ３が代表点として選択される。 Next, the pixel value P * of the target pixel is corrected with the error vector Vtot of the peripheral pixels. If the correction value is Pq,
Pq = P * + Vtot
It becomes. Then, the distance in the luminance / color difference space between Pq and representative points P0, P1, P2, and P3 is calculated, and the representative point that is the shortest distance is selected. In the illustrated example, P3 is selected as the representative point.
　このように重み付けされた周辺誤差を考慮して代表点を選択することによって、符号化誤差が拡散し、カラー画質の劣化を抑えることができる。上記では誤差拡散法を例に挙げたが、他にマトリクスディザ法等の２次元画像に用いられる階調量子化手法を用いることもできる。 By selecting the representative points in consideration of the weighted peripheral errors in this way, the coding error is diffused and the deterioration of the color image quality can be suppressed. In the above description, the error diffusion method is taken as an example. In addition, a gradation quantization method used for a two-dimensional image such as a matrix dither method can also be used.
　図３は、本実施例における画像符号化方法の流れを示すフローチャートである。
  ステップＳ１０１において原画像を入力し、ステップＳ１０２において画面を複数の処理領域（例えばブロック領域）に分割する。ステップＳ１０３において処理対象領域に移動する。 FIG. 3 is a flowchart showing the flow of the image encoding method in the present embodiment.
In step S101, an original image is input, and in step S102, the screen is divided into a plurality of processing areas (for example, block areas). In step S103, the process moves to the processing target area.
　ステップＳ１０４では、領域判定部１２により対象領域の色成分の分散値を求めて、その処理方式を判定する。すなわち、分散値が所定値より小さい場合はステップＳ１０５へ進み、輝度・色差空間における階調ベクトル量子化を行う。分散値が所定値より大きい場合はステップＳ１０８へ進み、画像符号化部１３により領域内を一般的な方式で画像符号化する。 In step S104, the region determination unit 12 obtains the dispersion value of the color component of the target region and determines the processing method. That is, when the variance value is smaller than the predetermined value, the process proceeds to step S105, and gradation vector quantization in the luminance / color difference space is performed. If the variance value is larger than the predetermined value, the process proceeds to step S108, and the image encoding unit 13 performs image encoding in the area by a general method.
　ステップＳ１０５では、代表点決定部１４により領域内の画素の輝度・色差情報から代表点を決定する。ステップＳ１０６では、階調ベクトル量子化部１５により代表点を用いて各画素の階調のベクトル量子化を行う。ステップＳ１０７では、代表点・量子化情報符号化部１６により代表点と各画素の量子化情報を符号化する。 In step S105, the representative point determination unit 14 determines a representative point from the luminance / color difference information of the pixels in the region. In step S106, the gradation vector quantization unit 15 performs vector quantization of the gradation of each pixel using the representative point. In step S107, the representative point / quantization information encoding unit 16 encodes the representative point and quantization information of each pixel.
　ステップＳ１０９では、前記Ｓ１０７またはＳ１０８で符号化したデータをメモリに蓄積する。ステップＳ１１０では全ての領域の処理を終了したかどうかを判定する。未処理の領域があれば前記ステップＳ１０３に戻り、次の処理対象領域に移動して符号化処理を続ける。 In step S109, the data encoded in S107 or S108 is stored in a memory. In step S110, it is determined whether or not the processing for all areas has been completed. If there is an unprocessed area, the process returns to step S103 to move to the next process target area and continue the encoding process.
　ステップＳ１１１では、符号統合部１７によりメモリに蓄積した各領域の符号化データを統合し符号化ストリームとして出力する。その際符号化ストリームには、各領域をいずれの符号化方式を用いたかを示すフラグを付与する。 In step S111, the encoded data stored in the memory by the code integration unit 17 is integrated and output as an encoded stream. At this time, a flag indicating which encoding method is used for each region is added to the encoded stream.
　図４は、本発明による画像復号化装置の一実施例を示す構成図である。画像復号化装置２は、符号化ストリームを入力し符号を分離する符号分離部２１と、画像を一般的な方式で復号化する画像復号化部２２と、代表点・量子化情報を復号化する代表点・量子化情報復号化部２３と、輝度・色差空間において各画素の階調ベクトルを復元する階調ベクトル復元部２４と、復元画像にフィルタ処理を行う高画質化フィルタ２５と、画像を合成する画像合成部２６と、画像を出力する画像出力部２７を備える。 FIG. 4 is a block diagram showing an embodiment of an image decoding apparatus according to the present invention. The image decoding device 2 receives a coded stream and separates a code, a code separation unit 21, an image decoding unit 22 that decodes an image by a general method, and a representative point / quantization information decoding unit The representative point / quantization information decoding unit 23, a gradation vector restoration unit 24 for restoring the gradation vector of each pixel in the luminance / chrominance space, a high image quality filter 25 for performing filtering on the restored image, and an image An image composition unit 26 for composition and an image output unit 27 for outputting an image are provided.
　各部の動作を説明する。符号分離部２１は符号化ストリームを入力し、付与されているフラグ等により、一般の方式によって符号化された領域（領域Ｂ）か、輝度・色差空間における階調ベクトル量子化によって符号化された領域（領域Ａ）かを判定する。画像復号化部２２は、領域Ｂと判定された符号化データについて一般的な方式で復号化し、復号画像を生成する。これには、画像符号化装置１の画像符号化部１３に対応する既存の画像復号化方式（例えばＨ．２６４／ＡＶＣ等）を用いる。 Explain the operation of each part. The code separation unit 21 receives an encoded stream, and is encoded by a gradation vector quantization in a region (region B) encoded by a general method or a luminance / chrominance space by a given flag or the like. It is determined whether it is an area (area A). The image decoding unit 22 decodes the encoded data determined to be the region B by a general method, and generates a decoded image. For this, an existing image decoding method (for example, H.264 / AVC) corresponding to the image encoding unit 13 of the image encoding device 1 is used.
　代表点・量子化情報復号化部２３は、領域Ａと判定された符号化データについて、代表点の輝度・色差の情報と、各画素にどの代表点を用いたかを示す情報（量子化情報）を復号化する。これには、画像符号化装置１の代表点・量子化情報符号化部１６に対応する復号化方式を用いる。階調ベクトル復元部２４は、代表点の輝度・色差の情報と各画素の量子化情報とを用いて、各画素の輝度・色差空間における階調ベクトルを復元する。 The representative point / quantization information decoding unit 23, with respect to the encoded data determined to be the region A, information on luminance / color difference of the representative point and information indicating which representative point is used for each pixel (quantization information) Is decrypted. For this, a decoding scheme corresponding to the representative point / quantization information encoding unit 16 of the image encoding device 1 is used. The gradation vector restoration unit 24 restores the gradation vector in the luminance / color difference space of each pixel using the luminance / color difference information of the representative point and the quantization information of each pixel.
　高画質化フィルタ２５では、復元された階調ベクトル（画素値）に対してフィルタ処理を行うことにより主観画質を向上させる。この処理は各色成分に対して独立に行ってもよいし、まとめて処理をしてもよい。最も簡単な方法は、各色成分に対してローパスフィルタをかけることである。復元画像においては各画素の値は代表点によって量子化されているが、周辺画素とのローパスフィルタ処理により、空間的な補間を行い画質を向上させることができる。他の方法としては、エッジの方向性を考慮したフィルタ処理や、ルックアップテーブル（ＬＵＴ）を用いたフィルタ処理等がある。 The image quality enhancement filter 25 improves the subjective image quality by performing filter processing on the restored gradation vector (pixel value). This process may be performed independently for each color component, or may be performed collectively. The simplest method is to apply a low pass filter to each color component. In the restored image, the value of each pixel is quantized by the representative point, but the image quality can be improved by performing spatial interpolation by low-pass filter processing with surrounding pixels. As other methods, there are filter processing in consideration of edge directionality, filter processing using a lookup table (LUT), and the like.
　画像合成部２６は、画像復号化部２２によって復号された画像と高画質化フィルタ２５によって出力された画像を合成し、画像出力部２４は合成された画像を出力する。 The image synthesis unit 26 synthesizes the image decoded by the image decoding unit 22 and the image output by the image quality enhancement filter 25, and the image output unit 24 outputs the synthesized image.
　以上のように本実施例の画像復号化装置２では、代表点を用いた階調のベクトル量子化により効率的に符号化された符号化ストリームを復号化し、高画質の復号画像を復元することが可能となる。 As described above, the image decoding apparatus 2 according to the present embodiment decodes an encoded stream efficiently encoded by vector quantization of gradations using representative points, and restores a high-quality decoded image. Is possible.
　図５Ａ～図５Ｃは、本実施例における階調ベクトル復元部２４と高画質化フィルタ２５から出力される画像の例を示す図である。 5A to 5C are diagrams illustrating examples of images output from the gradation vector restoration unit 24 and the image quality improving filter 25 in the present embodiment.
　図５Ａは復号化対象の原画像である。図５Ｂは、階調ベクトル復元部２４において代表点を用いて復元された画像である。この段階では、代表点の画素値のみで構成されるため、各画素を拡大するとバラバラの値がランダムに並んでいるようにも見える。 FIG. 5A shows an original image to be decoded. FIG. 5B is an image restored using the representative points in the gradation vector restoration unit 24. At this stage, since it is composed only of the pixel values of the representative points, it appears that when the pixels are enlarged, discrete values are arranged at random.
　図５Ｃは、高画質化フィルタ２５により処理した画像である。ここでは、ローパスフィルタあるいはエッジを考慮した適応的な非等方拡散フィルタ等を用いて周辺画素とのフィルタ処理を行い、図示するように高画質な画像を得ることができる。このようなフィルタ処理を組み合わせることによって、少ない代表点を用いた量子化にもかかわらず、高画質の復号画像を復元することができる。ここでのフィルタ処理による改善効果は、輝度・色差空間における階調ベクトル量子化と組み合わせることで特に有効となる。すなわち、輝度・色差空間での代表点を選択して３次元的に量子化したものであるから、復号画像において色ノイズの発生を抑えることができる。 FIG. 5C is an image processed by the image quality enhancement filter 25. Here, filter processing with peripheral pixels is performed using a low-pass filter or an adaptive anisotropic diffusion filter in consideration of edges, and a high-quality image can be obtained as shown in the figure. By combining such filtering processes, a high-quality decoded image can be restored despite quantization using a small number of representative points. The improvement effect by the filter processing here is particularly effective when combined with the gradation vector quantization in the luminance / color difference space. That is, since representative points in the luminance / color difference space are selected and quantized three-dimensionally, the generation of color noise in the decoded image can be suppressed.
　図６は、本実施例における画像復号化方法の流れを示すフローチャートである。
  ステップＳ２０１において符号化ストリームを入力し、ステップＳ２０２において画面の処理対象領域に移動する。 FIG. 6 is a flowchart showing the flow of the image decoding method in the present embodiment.
In step S201, an encoded stream is input, and in step S202, the process moves to a processing target area on the screen.
　ステップＳ２０３では、符号分離部２１により対象領域の符号化方式をストリームに付与されているフラグで判定する。フラグ＝「１」は階調ベクトル量子化による符号化の場合であり、ステップＳ２０４へ進む。フラグ＝「０」は一般の符号化方式による場合であり、ステップＳ２０７へ進み、画像復号化部２２により領域内を指定された方式で画像の復号化を行う。 In step S203, the code separation unit 21 determines the encoding method of the target area based on the flag added to the stream. Flag = “1” is the case of encoding by gradation vector quantization, and the process proceeds to step S204. The flag = “0” is based on a general encoding method, and the process proceeds to step S207, where the image decoding unit 22 performs image decoding using a method designated within the area.
　ステップＳ２０４では、代表点・量子化情報復号化部２３により符号化ストリームから
代表点と各画素の量子化情報を復号化する。ステップＳ２０５では、階調ベクトル復元部
２４により輝度・色差空間において代表点を用いて各画素の階調ベクトルの復元を行う。
ステップＳ２０６では、高画質化フィルタ２５により復元された画像にフィルタ処理を行
う。 In step S204, the representative point / quantization information decoding unit 23 decodes the representative point and the quantization information of each pixel from the encoded stream. In step S205, the gradation vector restoration unit 24 restores the gradation vector of each pixel using the representative point in the luminance / color difference space.
In step S206, the image restored by the image quality enhancement filter 25 is filtered.
　ステップＳ２０８では、画像合成部２６によりステップＳ２０６及びステップＳ２０７によって生成された画像を合成する。ステップＳ２０９では全ての領域の処理を終了したかどうかを判定する。未処理の領域があれば前記ステップＳ２０２に戻り、次の処理対象領域に移動し復号化処理を続ける。ステップＳ２１０では、画像出力部２７により合成された画像を出力する。 In step S208, the images generated in steps S206 and S207 are combined by the image combining unit 26. In step S209, it is determined whether or not the processing for all areas has been completed. If there is an unprocessed area, the process returns to step S202, moves to the next process target area, and continues the decoding process. In step S210, the image synthesized by the image output unit 27 is output.
　図７は、本発明による画像表示装置の一実施例を示す構成図である。本実施例の画像表示装置３は、階調ベクトル量子化によって符号化された符号化ストリームを入力するものとし、これを復号化してディスプレイ３３に画像表示するものである。その構成は、代表点・量子化情報を復号化する代表点・量子化情報復号化部３１と、輝度・色差空間において階調ベクトルを復元する階調ベクトル復元部３２とを有し、これらは実施例２（図４）における代表点・量子化情報復号化部２３と、階調ベクトル復元部２４と同様である。ただし本実施例では、図４における高画質化フィルタ２５を省略している。 FIG. 7 is a block diagram showing an embodiment of the image display device according to the present invention. The image display device 3 according to the present embodiment inputs an encoded stream encoded by gradation vector quantization, decodes this, and displays an image on the display 33. The configuration includes a representative point / quantization information decoding unit 31 that decodes representative point / quantization information, and a gradation vector restoration unit 32 that restores a gradation vector in the luminance / chrominance space. This is the same as the representative point / quantization information decoding unit 23 and the gradation vector restoration unit 24 in the second embodiment (FIG. 4). However, in this embodiment, the image quality enhancement filter 25 in FIG. 4 is omitted.
　本実施例の装置では、実施例２と同様に画像の復号化を行うが、復号画像に対してフィルタ処理を行わない。これは、人間の視野に対して画素が十分に小さく細かい場合、あるいは、遠くから眺めることによって相対的に画素が小さい場合には、人間の視覚には、復号画像に対してフィルタ作用が働くからである。すなわち、前記図５Ｂのように代表点のみで構成された画像が、図５Ｃのようなフィルタ後の画像のように見える。これは、例えば極少数の発色機能しか持たない電子ペーパーデバイスや、屋外の巨大ディスプレイ装置の場合に特に有効であり、階調のベクトル量子化による画像処理を行うことで、一般の量子化方法よりも高い画質で画像を表示することが可能となる。 In the apparatus of the present embodiment, the image is decoded as in the second embodiment, but the filtering process is not performed on the decoded image. This is because if the pixels are sufficiently small and fine with respect to the human visual field, or if the pixels are relatively small when viewed from a distance, the human vision will have a filtering effect on the decoded image. It is. That is, an image composed of only representative points as shown in FIG. 5B looks like an image after filtering as shown in FIG. 5C. This is particularly effective in the case of electronic paper devices that have a very small number of coloring functions and outdoor large display devices. In addition, it is possible to display an image with high image quality.
　以上のように各実施例で説明した画像符号化・復号化技術によれば、従来の符号化方式に比較し効率的な符号化が可能となり、特にカラー画質の劣化を抑えることで、高画質の復号画像を復元することが可能となる。 As described above, according to the image encoding / decoding technology described in each embodiment, it is possible to perform efficient encoding as compared with the conventional encoding method, and in particular, by suppressing deterioration of color image quality, It is possible to restore the decoded image.
　本発明の画像符号化装置、画像復号化装置、及びその方法は、画像記録装置、プレーヤ、携帯電話、携帯端末、デジタルカメラＴＶ、プロジェクタ、各種ディスプレイ、ゲーム機等の映像処理装置に広く適用することができる。 The image encoding device, the image decoding device, and the method thereof according to the present invention are widely applied to video processing devices such as an image recording device, a player, a mobile phone, a mobile terminal, a digital camera TV, a projector, various displays, and a game machine. be able to.
　１…画像符号化装置、２…画像復号化装置、３…画像表示装置、１１…画像入力部、１２…領域判定部、１３…画像符号化部、１４…代表点決定部、１５…階調ベクトル量子化部、１６…代表点・量子化情報符号化部、１７…符号統合部、２１…符号分離部、２２…画像復号化部、２３，３１…代表点・量子化情報復号化部、２４，３２…階調ベクトル復元部、２５…高画質化フィルタ、２６…画像合成部、２７…画像出力部、３３…ディスプレイ。 DESCRIPTION OF SYMBOLS 1 ... Image coding apparatus, 2 ... Image decoding apparatus, 3 ... Image display apparatus, 11 ... Image input part, 12 ... Area | region determination part, 13 ... Image coding part, 14 ... Representative point determination part, 15 ... Tone Vector quantization unit, 16 ... representative point / quantization information encoding unit, 17 ... code integration unit, 21 ... code separation unit, 22 ... image decoding unit, 23, 31 ... representative point / quantization information decoding unit, 24, 32 ... gradation vector restoration unit, 25 ... high image quality filter, 26 ... image composition unit, 27 ... image output unit, 33 ... display.

Claims (7)

  1. 　入力画像を符号化する画像符号化装置において、
    　輝度・色差空間における前記入力画像の画素値を代表する複数の代表点を決定する代表点決定部と、
    　輝度・色差空間における前記入力画像の処理対象画素と最短距離になる前記代表点を用いて、前記入力画像の階調のベクトル量子化を行う階調ベクトル量子化部と、
    　前記代表点の画素値の情報と前記入力画像の量子化情報とを符号化する代表点・量子化情報符号化部とを備え、
    　前記階調ベクトル量子化部は、前記処理対象画素の周辺画素をベクトル量子化した際の量子化誤差に前記処理対象画素との位置関係に応じた重み付けを行い、該量子化誤差の合計値が拡散するように前記処理対象画素の画素値を補正し、補正した画素値に対して最短距離になる前記代表点を選択することを特徴とする画像符号化装置。     In an image encoding device that encodes an input image,
    A representative point determination unit for determining a plurality of representative points representing pixel values of the input image in the luminance / color difference space;
    A gradation vector quantization unit that performs vector quantization of the gradation of the input image using the representative point that is the shortest distance from the processing target pixel of the input image in the luminance / color difference space;
    A representative point / quantization information encoding unit that encodes pixel value information of the representative point and quantization information of the input image;
    The gradation vector quantization unit weights the quantization error when the peripheral pixels of the processing target pixel are vector quantized according to the positional relationship with the processing target pixel, and the total value of the quantization error is calculated. An image encoding apparatus, wherein the pixel value of the processing target pixel is corrected so as to be diffused, and the representative point having the shortest distance with respect to the corrected pixel value is selected.
  2. 　請求項１に記載の画像符号化装置において、
    　前記入力画像の各処理対象領域における色成分の分散値を求め、該領域の符号化方式を判定する領域判定部を備え、
    　該領域判定部は、前記分散値が所定値より小さい領域に対し、前記した階調のベクトル量子化による画像の符号化を行なわせることを特徴とする画像符号化装置。     The image encoding device according to claim 1,
    An area determination unit that obtains a dispersion value of color components in each processing target area of the input image and determines an encoding method of the area,
    The area determination unit causes an image to be encoded by vector quantization of the gradation described above for an area where the variance value is smaller than a predetermined value.
  3. 　入力画像を符号化する画像符号化方法において、
    　輝度・色差空間における前記入力画像の画素値を代表する複数の代表点を決定する代表点決定ステップと、
    　輝度・色差空間における前記入力画像の処理対象画素と最短距離になる前記代表点を用いて、前記入力画像の階調のベクトル量子化を行う階調ベクトル量子化ステップと、
    　前記代表点の画素値の情報と前記入力画像の量子化情報とを符号化する代表点・量子化情報符号化ステップとを備え、
    　前記階調ベクトル量子化ステップでは、前記処理対象画素の周辺画素をベクトル量子化した際の量子化誤差に前記処理対象画素との位置関係に応じた重み付けを行い、該量子化誤差の合計値が拡散するように前記処理対象画素の画素値を補正し、補正した画素値に対して最短距離になる前記代表点を選択することを特徴とする画像符号化方法。     In an image encoding method for encoding an input image,
    A representative point determining step for determining a plurality of representative points representing pixel values of the input image in the luminance / color difference space;
    A gradation vector quantization step for performing vector quantization of gradation of the input image using the representative point that is the shortest distance from the processing target pixel of the input image in the luminance / color difference space;
    A representative point / quantization information encoding step for encoding pixel value information of the representative point and quantization information of the input image,
    In the gradation vector quantization step, the quantization error when the peripheral pixels of the processing target pixel are vector quantized is weighted according to the positional relationship with the processing target pixel, and a total value of the quantization errors is calculated. An image encoding method, wherein the pixel value of the processing target pixel is corrected so as to diffuse, and the representative point having the shortest distance with respect to the corrected pixel value is selected.
  4. 　請求項３に記載の画像符号化方法において、
    　前記入力画像の各処理対象領域における色成分の分散値を求め、該領域の符号化方式を判定する領域判定ステップを備え、
    　該領域判定ステップでは、前記分散値が所定値より小さい領域に対し、前記した階調のベクトル量子化による画像の符号化を行なわせることを特徴とする画像符号化方法。     The image encoding method according to claim 3, wherein
    An area determination step of obtaining a dispersion value of a color component in each processing target area of the input image and determining an encoding method of the area;
    In the region determining step, the image encoding method is characterized in that an image is encoded by vector quantization of the gradation described above for an area where the variance value is smaller than a predetermined value.
  5. 　入力した符号化ストリームを復号化する画像復号化装置において、
    　該符号化ストリームは、輝度・色差空間における代表点を用いて画像の階調をベクトル量子化することで画像を符号化したものであって、
    　前記符号化ストリームから、前記代表点の画素値の情報と画像の量子化情報とを復号化する代表点・量子化情報復号化部と、
    　前記代表点の画素値の情報と画像の量子化情報とを用いて、各画素の輝度・色差空間における階調ベクトルを復元する階調ベクトル復元部と、
    　復元された前記階調ベクトルに対してフィルタ処理を行う高画質化フィルタと、
    　を備えることを特徴とする画像復号化装置。     In an image decoding apparatus for decoding an input encoded stream,
    The encoded stream is obtained by encoding an image by vector quantization of the gradation of the image using a representative point in the luminance / color difference space,
    A representative point / quantization information decoding unit for decoding pixel value information and image quantization information of the representative point from the encoded stream;
    A gradation vector restoring unit that restores a gradation vector in the luminance / color difference space of each pixel using the pixel value information of the representative point and the quantization information of the image;
    An image quality enhancement filter that performs filtering on the restored gradation vector;
    An image decoding apparatus comprising:
  6. 　入力した符号化ストリームを復号化する画像復号化方法において、
    　該符号化ストリームは、輝度・色差空間における代表点を用いて画像の階調をベクトル量子化することで画像を符号化したものであって、
    　前記符号化ストリームから、前記代表点の画素値の情報と画像の量子化情報とを復号化する代表点・量子化情報復号化ステップと、
    　前記代表点の画素値の情報と画像の量子化情報とを用いて、各画素の輝度・色差空間における階調ベクトルを復元する階調ベクトル復元ステップと、
    　復元された前記階調ベクトルに対してフィルタ処理を行うフィルタ処理ステップと、
    　を備えることを特徴とする画像復号化方法。     In an image decoding method for decoding an input encoded stream,
    The encoded stream is obtained by encoding an image by vector quantization of the gradation of the image using a representative point in the luminance / color difference space,
    A representative point / quantization information decoding step for decoding pixel value information and image quantization information of the representative point from the encoded stream;
    A gradation vector restoring step for restoring a gradation vector in the luminance / color difference space of each pixel using the pixel value information of the representative point and the quantization information of the image;
    A filtering process step for performing a filtering process on the restored gradation vector;
    An image decoding method comprising:
  7. 　入力した符号化ストリームを復号化して表示する画像表示装置において、
    　該符号化ストリームは、輝度・色差空間における代表点を用いて画像の階調をベクトル量子化することで画像を符号化したものであって、
    　前記符号化ストリームから、前記代表点の画素値の情報と画像の量子化情報とを復号化する代表点・量子化情報復号化部と、
    　前記代表点の画素値の情報と画像の量子化情報とを用いて、各画素の輝度・色差空間における階調ベクトルを復元する階調ベクトル復元部と、
    　前記復元した階調ベクトルに従い画像を表示するディスプレイと、
    　を備えることを特徴とする画像表示装置。     In an image display device that decodes and displays an input encoded stream,
    The encoded stream is obtained by encoding an image by vector quantization of the gradation of the image using a representative point in the luminance / color difference space,
    A representative point / quantization information decoding unit for decoding pixel value information and image quantization information of the representative point from the encoded stream;
    A gradation vector restoring unit that restores a gradation vector in the luminance / color difference space of each pixel using the pixel value information of the representative point and the quantization information of the image;
    A display for displaying an image according to the restored gradation vector;
    An image display device comprising:
PCT/JP2010/003341 2009-06-15 2010-05-18 Image encoding device, image encoding method, image decoding device, image decoding method, and image display device WO2010146769A1 (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
JP2011519503A JP5342645B2 (en) 2009-06-15 2010-05-18 Image coding apparatus and image coding method

Applications Claiming Priority (2)

Application Number Priority Date Filing Date Title
JP2009142262 2009-06-15
JP2009-142262 2009-06-15

Publications (1)

Publication Number Publication Date
WO2010146769A1 true WO2010146769A1 (en) 2010-12-23

Family

ID=43356104

Family Applications (1)

Application Number Title Priority Date Filing Date
PCT/JP2010/003341 WO2010146769A1 (en) 2009-06-15 2010-05-18 Image encoding device, image encoding method, image decoding device, image decoding method, and image display device

Country Status (2)

Country Link
JP (1) JP5342645B2 (en)
WO (1) WO2010146769A1 (en)

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS63158972A (en) * 1986-12-23 1988-07-01 Matsushita Electric Works Ltd Image compression system
JPH02279080A (en) * 1989-04-20 1990-11-15 Oki Electric Ind Co Ltd Picture coder and decoder
JPH04188952A (en) * 1990-11-22 1992-07-07 Canon Inc Color picture communication system
JPH118848A (en) * 1997-06-19 1999-01-12 Matsushita Electric Ind Co Ltd Image encoding method and device therefor
JP2003131652A (en) * 2001-07-06 2003-05-09 Eastman Kodak Co Method for representing digital color image using set of palette color based on detected important color
JP2005117502A (en) * 2003-10-09 2005-04-28 Sony Corp Encoding sevice and encoding method, decoding device and decoding method, program and recording medium

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
JPS63158972A (en) * 1986-12-23 1988-07-01 Matsushita Electric Works Ltd Image compression system
JPH02279080A (en) * 1989-04-20 1990-11-15 Oki Electric Ind Co Ltd Picture coder and decoder
JPH04188952A (en) * 1990-11-22 1992-07-07 Canon Inc Color picture communication system
JPH118848A (en) * 1997-06-19 1999-01-12 Matsushita Electric Ind Co Ltd Image encoding method and device therefor
JP2003131652A (en) * 2001-07-06 2003-05-09 Eastman Kodak Co Method for representing digital color image using set of palette color based on detected important color
JP2005117502A (en) * 2003-10-09 2005-04-28 Sony Corp Encoding sevice and encoding method, decoding device and decoding method, program and recording medium

Also Published As

Publication number Publication date
JP5342645B2 (en) 2013-11-13
JPWO2010146769A1 (en) 2012-11-29

Similar Documents

Publication Publication Date Title
US8594188B2 (en) High precision encoding and decoding of video images
JP5868925B2 (en) Method and apparatus for post-processing assisted by an encoder
JP5989934B2 (en) Encoding perceptual quantized video content in multi-layer VDR coding
US6697521B2 (en) Method and system for achieving coding gains in wavelet-based image codecs
JP5290171B2 (en) Method and apparatus for post-processing assisted by an encoder
US20210067785A1 (en) Video encoding rate control for intra and scene change frames using machine learning
KR101089394B1 (en) Visual processing device, visual processing method, visual processing program, and semiconductor device
US20080123979A1 (en) Method and system for digital image contour removal (dcr)
JP2012199963A (en) Variance based adaptive block size dct image compression
WO2002102050A2 (en) System and method for enhancing digital video
JP4360416B2 (en) Image compression method, image compression apparatus, and program
JP2009153128A (en) Selective chrominance decimation for digital images
JP2011175085A (en) Display driving circuit
JP4641784B2 (en) Gradation conversion processing device, gradation conversion processing method, image display device, television, portable information terminal, camera, integrated circuit, and image processing program
CN109874012B (en) Video coding method, encoder, electronic device and medium
JP2004528791A (en) Inter-frame encoding method and apparatus
CN111556320A (en) Data processing system
JP4552400B2 (en) Image display device, image display method, and image display program
JP5342645B2 (en) Image coding apparatus and image coding method
US20230412808A1 (en) Determining adaptive quantization matrices using machine learning for video coding
CN116248895B (en) Video cloud transcoding method and system for virtual reality panorama roaming
US20210136378A1 (en) Adaptive quality boosting for low latency video coding
EP3099073A1 (en) Method and device of encoding/decoding a hdr and a sdr picture in/from a scalable bitstream
JP5181059B2 (en) Image encoding device, image decoding device, image encoding method, and image decoding method
CN115442636A (en) Live video stream conversion method, device, equipment and storage medium

Legal Events

Date Code Title Description
121 Ep: the epo has been informed by wipo that ep was designated in this application

Ref document number: 10789154

Country of ref document: EP

Kind code of ref document: A1

WWE Wipo information: entry into national phase

Ref document number: 2011519503

Country of ref document: JP

NENP Non-entry into the national phase

Ref country code: DE

122 Ep: pct application non-entry in european phase

Ref document number: 10789154

Country of ref document: EP

Kind code of ref document: A1