By Tina Nichols,2014-06-25 12:28
7 views 0
DCT the Achilles heel of zero-tree coding. (2) layer-type DCT zerotree coding The algorithm for the DCT transform the image will be low-frequency ...



    Digital image compression technology Research and Development

    Tian Yong 1, DING Xue-Jun 2

    (1. Northeastern University, embedded systems Neusoft Institute of

    Information Engineering, Liaoning, Dalian 116023; 2. Northeast University of

    Finance and Information Engineering, Liaoning, Dalian 116023)

    Abstract: Digital image compression technology for digital image information

    on the network for rapid transfer and real-time processing is of great

    significance. This article describes several of the most important of the current


    Compression algorithms: JPEG, JPEG2000, fractal image compression and

    wavelet transform image compression, summed up the advantages and

    disadvantages of their development prospects. A brief introduction and then

    arbitrarily shaped visual object coding Algorithm for the status quo, and pointed out that this algorithm is a high

    compression ratio resulting image compression algorithms.

    Keywords: JPEG; JPEG2000; fractal image compression; wavelet transform;

    arbitrarily shaped visual object coding CLC number: TP3 Abstract: A Article ID :1672-545X (2007) 04-0072-04 With

    the multimedia technology and communication technology continues to evolve,

    multimedia entertainment, the letter

    Interest rates have kept the information highway data storage and

    transmission should be set higher

    Demand, but also to the limited bandwidth available to a severe test, especially

    those with large data

    The amount of digital image communication, more difficult to transport and

    storage, which greatly restricted the image pass The development of the letter, so the image compression technology has been

    more and more attention. Image compression The purpose is to shrink the original image of a larger less bytes exhausted

    expression and transmission,

    And asked to recover a better quality image. The use of image compression

    that can reduce the map

    Such as the burden of storage and transmission, so that the image on the

    network to achieve fast transfer and real-time Office Li.

    Image compression techniques can be traced back to 1948, raised the number

    of television signals

    Word-oriented, and today has 50 years of history [1]. During this period there

    have been many

    Kinds of image compression coding method, especially in the late 20th century,

    80 years later,

Because the wavelet transform theory, fractal theory, artificial neural network

    theory, visual simulation of Li

    On the establishment of image compression technology has been an

    unprecedented development, in which fractal image Image compression and wavelet image compression is the current hot

    research. In this paper, present, the most widely Pan-use image compression algorithms are reviewed and discussed their

    advantages and disadvantages, as well as hair Show prospect.

    1 JPEG compression

    Responsible for developing the still image compression standard "Joint

    Photographic Expert Group" (Joint

    Photographic Expert Group, referred to as JPEG), was formed in January


    Adaptive DCT-based JPEG technical specifications of the first draft, followed

    by several

    Changes to the 1991 formation of the draft international standard ISO10918,

    and after one year into the

    As an international standard, called JPEG standard. 1.1 JPEG compression principle and characteristics of JPEG algorithm in the first block of the image processing, generally divided

    into each other without re -

    The size of stacked blocks, and then every one to two-dimensional discrete

    cosine transform (DCT). Variable

    After changing coefficients are not relevant, and the coefficient matrix of the

    energy concentrated in the low-frequency area, the root Quantization table, according to quantify, quantify the results of retained

    low-frequency part of the coefficient, remove the The high-frequency part of the coefficients. The quantized coefficients by

    Zigzag scan re-organization, Ran

    After Huffman coding. JPEG features are as follows: Advantages: (1) the formation of international standards; (2) has a mid-range

    and high bit-rate on the

    Good image quality.

    Disadvantages: (1) Because of the image block, at high compression ratio

    when you have a serious

    Blocking effects; (2) factor to quantify is the lossy compression; (3) the

    compression ratio is not high, small

    At 50 [2].

    JPEG compressed image box effect occurs because: Under normal

    circumstances the image

    The signal is highly non-stationary, it is difficult to characterize the process of

    using Gauss, and the image

    Some of the mutant structure, such as the edge information than the image

     smoothness important, with the cosine-based

    Non-linear approximation for image signals the result is not optimal [3].

    1.2 JPEG compression status and prospects of the study [2]

    JPEG at high compression ratio for cases arising under the box effect,

    decompresses the image than the Poor, in recent years made a lot of improvement methods, the most effective is

    the following two ways:

    (1) DCT zerotree coding

    DCT block zero-tree coding the DCT coefficients in the composition of log2N

    sub-zone, Ran

    After the zero-tree coding scheme used to encode. Compression ratio in the

    same circumstances, the

    PSNR value higher than the EZW. However, in the case of high compression

    ratio, the box effect is still

    DCT the Achilles heel of zero-tree coding. (2) layer-type DCT zerotree coding The algorithm for the DCT transform the image will be low-frequency blocks

    together, doing anti -

    DCT; pairs of newly acquired images do the same transformation, and so it

    goes, until satisfied

    Request until the. Then Layered DCT transform and zero-tree arrangement of

    the coefficients over zero

    Tree coding.

    JPEG compression one of the biggest problem is that at high compression

    ratio when you have a serious Blocking effects, so future research should be focused on solving the resulting

    DCT transform

    Blocking effects, taking into account the human visual system with the

    combination of compression. 2 JEPG2000 compression

    JPEG2000 is a ISO / IEC JTCISC29 standards group responsible for


    The new still image compression standard. One of the biggest improvements

    is that it uses wavelet transform on behalf of the

    For the cosine transform. In March 2000 the Tokyo conference to identify a

    color still image

    Encoding a new generation of image compression standard-JPEG2000 coding


    2.1 JPEG2000 compression principle and characteristics of

    JPEG2000 Codec encoder and decoder block diagram in Figure 1 [4].

    Encoding process is mainly divided into the following processes:

    pre-processing, core processing and bit Stream organization. Pretreatment section includes image segmentation, DC

    level (DC) displacement and sub -

Received date: 2007 - 01 - 16

    Author: Takada (1975 -), male, Master, Lecturer, major research interests

    include digital image processing, integrated circuit design, etc.; DING Xue-jun

    (1978 -), female, master's degree, teaching assistants, the main research

    direction for information

    And network security, digital image processing. The amount of change. Core

    processing in part by the discrete wavelet transform, quantization and entropy

    coding form. Bit

    Stream part of the organization include zoning, code blocks, layers and

    packages organization.

    JPEG2000 format image compression ratio can be based on the current JPEG

    Further increased by 10% to 30%, and the compressed image is even more

    delicate smooth. For

    The current JPEG standard, in the same compressed stream can not provide

    lossy and

    Lossless compression, while in JPEG2000 system, by selecting the

    parameters, be able to image

    For lossy and lossless compression. Now the network is based on JPEG

    image download

    "Block" transmission, while the JPEG2000 image format support for

    progressive transmission, which use the Households do not receive the entire image compression bit stream. Since

    JPEG2000 uses wavelet technology Surgery may be random access to some image region of interest (ROI) of the

    compressed stream, right

    Compressed image data transmission, filtering and other operations [4].

    Figure 1 JPEG2000 compression and decompression of the overall process

    2.2 JPEG2000 compression prospects JPEG2000 standard applies to a variety of image compression coding. Its

    application field

    Will include the Internet, fax, printing, remote sensing, mobile communications,

    medical, digital libraries

    And e-commerce [5]. JPEG2000 image compression standard will become the

    21st century, the main

    Flow still image compression standard. 3 Wavelet Image Compression 3.1 Principle of Wavelet Image Compression Wavelet Transform for image coding basic idea is that the image according to


    Tower fast wavelet transform algorithm for multi-resolution decomposition. The

    specific process is:

    First, the image multi-level wavelet decomposition, then the amount of the

    wavelet coefficients of each layer , And re-pairs of the quantized coefficients encoded. Wavelet image

     compression is the current image compression

    One of the hot shrinkage has been formed based on wavelet transform of

    international compression standards, such as the MPEG-4 standard, and as mentioned above the JPEG2000 standard [2].

    3.2 Wavelet Image Compression Status and Prospects At present the three highest levels of wavelet image coding are embedded

    wavelet zero -

    Tree Image Coding (EZW), hierarchical tree distribution of the sample image

    coding (SPIHT) and

    Expand image compression coding (EBCOT). (1) EZW coder [6]

    In 1993, Shapiro introduced a wavelet "zero-tree" concept, by defining

    POS, NEG, IZ and ZTR symbols are four kinds of recursive spatial wavelet tree

    coding, there are

    Effective to remove the encoding of high-frequency coefficients, which greatly

    improved the coding of wavelet coefficients Efficiency. This algorithm is used to quantify and embedded progressive

    coding mode, the algorithm complexity Low. EZW algorithm for breaking the long-term deep faith in the field of

    information processing criteria: highly efficient compression

    Reduction encoder must pass the highly complex algorithms to get, so EZW


    Data compression device in the history of a landmark decision.

    (2) SPIHT encoder [7]

    By Said and Pearlman proposed a collection of hierarchical wavelet tree

    segmentation algorithm

    (SPIHT) is the use of space partitioning tree hierarchical approach to

    effectively reduce the bit-plane

    Coded symbol set size. EZW, compared with, SPIHT algorithm is constructed

    by two kinds of non -

    The same type of space zero-tree, to make better use of the wavelet

    coefficients of the amplitude attenuation. Like with the EZW encoder, SPIHT encoder algorithm complexity low, the


    Is also an embedded bit stream, but the encoder performance than EZW

    greatly improved.

    (3) EBCOT encoder [8]

    Optimized cut-off point of the embedded block coding (EBCOT) first wavelet


    Each sub-band is divided into a relatively independent of the code block, and

    then use a hierarchical optimization

    Truncation algorithm to the code block is encoded to generate compressed

    stream, the results of image compression Reduction stream is not only scalable but also has resolution of SNR can be

Extension, you can also support the random image is stored. In comparison,

    EBCOT algorithm complexity compared with EZW and SPIHT are referred to

    With its high compression performance than SPIHT increased slightly.

    Wavelet image compression is considered the most promising One of the image compression algorithm. Wavelet image compression

    research has focused

    Wavelet coefficients in the coding issues. In later work, Should fully consider the human visual system to further improve the

    compression ratio,

    To improve the image quality. And consider the wavelet transform and other


    Method combination. For example, with the combination of fractal image

    compression is the current

    A research focus [2].

    4 Fractal Image Compression

    In 1988, Barnsley proved by experiments fractal image compression ratio can


    Classical image coding technology, high compression ratio of several orders of

    magnitude. In 1990, Barnsley

    Students AEJacquin making a partial iterated function system theory, will bring

    the use of fractal

    In image compression on the computer automatically possible. 4.1 The principle of fractal image compression Fractal compression using mainly self-similar characteristics, through the

    iterated function system (Iterated

    Function System, IFS) to achieve. The theory is based on iterated function


    Theorem and the collage theorem.

    Fractal image compression the original image is divided into several

    sub-images, and then each

    Sub-image corresponds to an iterated function, sub-images to iterated function

    storage, iterative letter

    The number of the more simple, the greater the compression ratio. The same

    decoding as long as the transfer out of each sub-map Like the corresponding iteration function iteration, it can be restored out of the

    original sub-image, from the

    Obtained the original image [9].

    4.2 several major fractal image coding techniques [9] With the fractal image compression technology, an increasing number of

    algorithms have been proposed,

    Based on the different fractal characteristics can be divided into several major

    coding method.

    (1) The size of encoding method

    Size coding method is based on fractal geometry, the use of small-scale

     irregular metric

    Curve length method, similar to the traditional sub-sampling and interpolation

    methods, the main non -

    Scale with the encoding method is that the introduction of the idea of fractal,

    scale with the image

    The complexity of the various components of the different change.

    (2) The method of iterated function system Iterated function system method is the most studied, the most widely used as a

    sub -

    Shaped compression technology, it is a patchwork of human-computer

    interaction technology, it is based on the natural world map Like in the prevailing global and local autocorrelation features, search for such

    auto-correlation mapping

    Shot between the expression, that is, affine transformation, and by storing a

    small amount compared with the original image data Affine coefficients to achieve the purpose of compression. If you find a simple

    affine transformation Effective, then the iterated function system can achieve

    very high compression ratio.

    (3) A-E-Jacquin fractal program

    A-E-Jacquin fractal program is an automated block-based fractal

    Image compression scheme, it is also looking for a mapping process, but are

    looking for right

    As the domain is divided into blocks of the image after the partial and local

    interests. In this scenario,

    There are still some redundancy can be removed, and its decoding the image

    there is a clear

    Blocking effects.

    4.3 The prospects for fractal image compression [2] Although the fractal image compression in the field of image compression is

    not dominant, but the

    Fractal Image Compression take into account both local and partial, but also

    take into account the relevant part and the whole Nature, suitable for self-similar or self-affine image compression, but in nature

    there is a large number of

    Self-similarity or self-affine geometry, so its scope is broad. 5 other compression algorithms

    In addition to these several common image compression methods, there are:

    NNT (number theory

    Transform) compression, the compression method based on neural networks,

    Hibert scanned image compression side France, adaptive polyphase subband compression method, etc. This is not to

    repeat. Following is a brief

    In recent years, several arbitrarily shaped texture coding algorithm [10 ~ 13].

    (1) The shape-adaptive DCT (SA-DCT) algorithm

SA-DCT to an arbitrarily shaped visual objects into the image blocks, each


    To DCT transform, which implements a similar shape-adaptive Gilge DCT [10,


    Transform an effective transformation, however, it Gilge DCT transform lower

    complexity. But

    SA-DCT also has shortcomings, it is the pixel onto one side of the border with

    the rectangle relative to flat

    Qi, so some spatial correlation may be missing, so further out DCT transform,

    There is a greater distortion of the [11, 14, 15].

    (2) Egger method

    Egger and others [16, 17] proposed a wavelet applied to objects of arbitrary

    shape change

    For the program. In this scenario, the first visible line of pixels onto the object

    and bounding box

    The right boundary phase flush position, then the usefulness of each line pixel

    wavelet changes

    Change, followed by a further direction of the wavelet transform. This program,

    take advantage of a small

    The local characteristics of wavelet transform. However, the program also has

    its problems, such as might lead to Play an important part of the high-frequency part of the merger with the

    boundary, can not guarantee that the distribution coefficient of each of the

    The same inter-correct phase, as well as the second direction may cause the

    non-wavelet decomposition

    Continuous and so on.

    (3) The shape-adaptive discrete wavelet transform (SA-DWT)

    Li et al proposed a novel encoding of arbitrary shape of the object, SA-DWT

    Coding [18 ~ 22]. The technology includes the SA-DWT and zerotree entropy

    coding extension

    (ZTE), and embedded wavelet coding (EZW). SA-DWT is characterized by:


    SA-DWT coefficients after the count of visual objects of arbitrary shape with

    the original number of pixels The same; wavelet transform of the spatial correlation, regional attributes, and

    sub-band between the Zi Xiang Similarity, in the SA-DWT can be well manifested in; for the rectangular area,

    SA-DWT and the same as the traditional wavelet transform. SA-DWT coding

    technology to achieve

    Have been new multimedia coding standard MPEG-4's for any shape of a

    static pattern

    Rationale used by the encoding. In future work, you can fully utilize the human visual system to the image edge

    Edge of some of the more sensitive features, try to image objects of interest to

     split up,

    Some of its edges, internal textures and objects other than some background

    section according to different Compression ratio of compression, so that compressed images can achieve

    greater compression ratio and more

    Add to facilitate transmission. 6 Summary

    Image compression technology research for decades and has made great

    achievements, but there are Much to be desired, it is worth further examination. Wavelet image

    compression and fractal image compression

    Reduction is currently a hot research, but both also have their own

    shortcomings, in the future work, Should be with the human visual properties. In short, image compression is a

    very developed

    Promising area of research, a breakthrough in this area of life for our

    information and communication To the development of far-reaching implications.




















    Tian Qing. Image compression techniques [J]. Police Technology, 2002, (1): 30

    - 31.

    ZHANG Hai-yan, Wang Mu, et al. Image compression technique [J]. Journal of

    System Simulation, 2002, 14 (7): 831 - 835.

    Zhang Zongping, Liu Guizhong. Wavelet-based video image compression

    research [J]. Journal of Electronics,

2002, 30 (6): 883 - 889.

    Zhou Ning, Tang Xiaojun, Xu Park. JPEG2000 image compression standard

    and its key algorithm [J].

    Modern electronic technology, 2002, (12): 1 - 5.

    WU Yonghui, YU Jian-xin. JPEG2000 image compression algorithm overview

    and network application prospects [J].

    Computer Engineering, 2003, 29 (3): 7 - 10.

    J MShaprio. Embedded image coding using zerotree of wavelet coefficients [ J]. IEEE Trans. On Signal Processing, 1993, 41 (12): 3445 - 3462. A Said, WA Pearlman. A new fast and efficient image codec based on set partitioning in hierarchical trees [J]. IEEE Trans. on Circuits and Systems for Video Tech. 1996, 6 (3): 243 - 250.

    D Taubman. High performance scalable image compression with EBCOT [ J]. IEEE Transactions on Image Processing, 2000, 9 (7): 1158 -- 1170.

    Xulin Jing, MENG Li-min, Jianjun. Wavelet image compression with branches

    in comparison and applications.

    China's cable TV, 2003, 03/04: 26 - 29.

    M Gilge, T Engelhardt, R Mehlan. Coding of arbitrarily shaped image segments based on a generalized orthogonal transform [J]. Signal Processing: Image Commun., 1989, 1 (10): 153-180.

    T Sikora, B Makai. Shape-adaptive DCT for generic coding of video [J]. IEEE Trans. Circuits Syst. Video Technol., 1995, 5 (1): 59-62. T Sikora, S Bauer, B Makai. Efficiency of shape-adaptive 2 - D transforms for coding of arbitrarily shaped image segments [J]. IEEE Trans. Circuits Syst. Video Technol., 1995, 5 (3): 254-258.

    E Jensen, K Rijk, et al. Coding of arbitrarily shaped image segments [C]. Proc. Workshop Image Analysis and Synthesis in Image Coding, Berlin, Germany, 1994: E2.1-E2.4.

    M Bi, SH Ong, YH Ang. Comment on "Shape-adaptive DCT for generic coding of video "[J]. IEEE Trans. Circuits Syst. Video Technol., 1996, 6 (6): 686-688.

    P Kauff, K Schuur. Shape-adaptive DCT with block-based DC separation and Delta DC correction [J]. IEEE Trans. Circuits Syst. Video Technol., 1998, 8 (3): 237-242.

    O Egger, P Fleury, T Ebrahimi. Shape-adaptive wavelet transform for zerotree coding [C]. Proc. Eur. Workshop Image Analysis and Coding for TV, HDTV and Multimedia Application, Rennes, France, 1996: 201 -- 208.

    O Egger. Region representation using nonlinear techniques with applications to image and video coding [D]. Ph.D. dissertation, Swiss Federal Institute of Technology (EPFL), Lausanne, Switzerland, 1997. S Li, W Li, et al. Shape adaptive vector wavelet coding of arbitrarily shaped texture [S]. ISO / IEC JTC/SC29/WG11, MPEG-96 - m1027,

Report this document

For any questions or suggestions please email