毕业设计_开题报告_图像压缩技术的研究及

-河南城建学院本科毕业设计(论文)开题报告 题目： 图像压缩算法的研究课 题 类 型： 论文学 生 姓 名： 吕建辉 学 号： 093410130专 业 班 级： 0934101系 别： 电气系指 导 教 师： 石磊开 题 时 间： 2014年3月 2014 年 04 月 10日一、毕业设计内容及研究意义设计的内容：本论文的主要研究内容是图像压缩技术。具体框架是首先介绍了图像压缩的基本原理以及其相关压缩方法分类等理论知识，并且说明了对图像进行压缩的必要性与重要性，然后针对目前图像压缩现状和发展趋势，着重介绍了小波变换，并以其为基础来进行数字图像的压缩处理，这也许会成为图像数据压缩的主要技术之一。接着又根据相关知识编写了一些简单的图像处理程序，对前面的理论进行试验、分析、论证。最后，对整篇论文进行总结，发现自身研究的不足，并展望其未来发展前景研究意义：图像信息给人们以直观、生动的形象，正成为人们获取外部信息的重要途径。然而，数字图像具有极大的数据量，在目前的计算机系统的条件下，要想实时处理，若图像信息不经过压缩，则会占用信道宽，是传输成本变得昂贵，传输速率变慢。这对图像存储、传输及使用都非常不利，同时也阻碍了人们对图像的有效获取和使用。另外，伴随着计算机科学技术的发展，图像压缩技术在通信系统和多媒体系统中的重要性也越来越高，在我们的学习、生产、生活以及国防事物中等的作用越来越显著。为此，人们给予了图像压缩技术广泛的关注，如何用尽量少的数据量来表示图像信息，即对图像进行压缩，越来越成为图像研究领域的重点课题。二、毕业设计研究现状和发展趋势研究现状：第一代图像压缩编码的研究工作是从上个世纪50年代提出电视信号数字后开始的，至今己有60多年的历史。主要是基于信息论的编码方法，压缩比小。1966年J.B.Neal对比分析了差分编码调制(DPCM)和脉冲编码调制(PCM)并提出了用于电视的实验数据，1969行了线性预测编码的实际实验。同年举行首届图像编码会(PictureCodiSymP0sium)，在这次会议之后，图像压缩编码算法的研究有了很大进展。由于DCT压缩算法具有编码效果较好、运算复杂度适中等优点，目前己经成为国际图像编准(JPEG)的核心算法。为了克服第一代图像压缩编码存在的压缩比小、图像复原质量不理想等1985年Kunt等人充分利用人眼视觉特性提出了第二代图像压缩编码的概念。上世纪80年代中后期，人们相继提出了在多分辨率下表示图像的方案，主要方子带压缩编码、金字塔压缩编码等。这些方法均在不同程度上有如下优点:多分辨率的信号表示有利于图形信号的渐输，不同分辨率的信号占用不同的频带，便于引入视觉特性。1987年，Mallat次巧妙地将计算机视觉领域内的多尺度分析思想引入到小波变换中，统一了在之前各种小波的构造方法之后，他又研究了小波变换的离散形式，并将相应的法应用于图像的分解与重构中，为随后的小波图像压缩编码奠定了理论基础。1988年Barnsly和S1an共同提出了分形图像编码压缩方案，之后，各国学者提出各种各样的改进方法，从而掀起了分形图像编码的新高潮。但由于在分形压缩编码过程中，运算量大，从而造成编码时间过长，且提高压缩比同减小失真度之间的矛盾始终存在，从而局限了它的实用性。上个世纪90年代后，又取得了一系列图像压缩编码研究的阶段性新成果，基于零树的编码法首先由A.5.Lewis和G.Knoes提出，其特点是根据小波系数在同方向子带中的相似性，即若一个小波系数较小，则很可能高一级(频率更高)同方向子带中相应位置的小波系数也较小，利用一种称为小波树的树形结构来组织小波系数，使其能方便地去除频域和空间域中的相关性。接着Shapir结合比特平面编码方法设计了更好的零树编码方法，Shapri提出的嵌入式零树小波算法是迄今为止最有效的方法，它有效地利用了小波系数的特性，实现了图像的可分级编码，但是不同程度地存在算法时间复杂度和空间复杂度过高的弱点。目前，小波变换的图像压缩编码算法已成为图像压缩研究领域的一个主要方向，基于小波变换的图像编码技术正逐步显示出它的优越性，其中小波变换己被JPEG-2000国际标准采用。发展趋势：随着数字化技术的迅速发展，数字图像也被越来越多的运用在我们的日常学习、生活、工作当中，图像压缩技术将会非常的重要。从国际数据压缩技术的发展尤其是MPEG的发展可以看出，基于内容的图像压缩编码方法是未来编码的发展趋势。它不仅能满足进一步获得更大的图像数据压缩比的要求，而且能够实现人机对话的功能。另外，任意形状物体的模型建立的关键问题还没有解决，这严重影响其应用的广泛性。因此，图像编码将朝着多模式和跨模式的方向发展。通过元数据进行编码也是今后编码的发展方向。元数据是指详细的描述音/视频信息的基本元素，利用元数据来描述音视频对象的同时也就完成了编码，因为此时编码的对象是图像的一种描述而不再是图像本身。从另一个角度来说，进一步提高压缩比，提高码流的附属功能（码流内容的可访问性、抗误码能力、可伸缩性等）也将是未来的编码的两个发展方向。三、毕业设计研究方案及工作计划1、研究方案此次论文撰写拟从图像数据压缩的必要性出发，阐述本课题的研究义，然后简单介绍了图像压缩编码技术的发展现状，最后对本文的主要工作以及各章的内容编排做了说明。接下来对数字图像压缩的必要性、重要性以及现今国内外对于这一技术领域的研究现状介绍开始，进而系统并简明的述说图像压缩理论知识。紧接着将对为何要选用小波变换为基础，来进行数字图像的压缩处理进行具体且详细的论述，并且介绍了小波变换对静止的图像进行编码的技术，阐述了Matlab算法，最后附上相关程序，通过实验来验证本论文所选方法的优越性。最后是对此次论文的总结。具体有以下几方面的工作：（1）、介绍图像及图像压缩的相关知识，弄清楚图像定义、基本概念及压缩编码的基本过程，并且重点讨论几种传统的、在图像压缩中被广泛采用的编码方法及其基本原理。（2）、通过分析比较小波变换和传统变换方法，发现小波变换在图像压缩技术领域的优越性。同时，从小波基的选择、原始信号的边界延拓、小波系数的量化、熵编码、小波分解/重构级数来逐步分析小波编码的基本思想、理论。（3）、量化和编码间的关系是小波变换压缩的关键所在。本论文中将用EZW与SPIHT算法来处理这种关系，即研究该算法的主要思想、实现步骤，并阐述、编写基于Matlab软件的相关程序，通过程序的运行来验证所用方法的优越性。2、重点和难点此次论文的重难点是，对为何要选用小波变换为基础，来进行数字图像的压缩处理进行：（1）要介绍小波变换的背景及基本理论、函数，其中包括对于传统变换方法的局限性（对瞬态和局部信号分量的分析、时频和空频局部化）的论述，小波变换的分类及其它；（2）分析小波参数的选择及其对图像压缩前后的质量的影响，以及怎样利用小波系数的分布特点和小波图像特点，对系数进行组织和编码，实现数字图像的压缩；（3）介绍两种经典的基于小波变换的图像压缩编码算法即EZW算法和SPIHT算法，并研究其主要思想、具体实现步骤，同时，认真分析各种算法的优点和缺点试图寻找可以改进的地方。最后阐述了Matlab算法，利用小波变换理论进行相关程序的编写，通过实验来验证所选方法的优越性。3、工作计划学生姓名吕建辉专业电子信息工程起止日期（日/月）周次内 容 进 程备 注3.01-3.071接受设计的课题，查找相关参考文献和资料。3.08-3.142熟悉设计的课题，查阅、整理参考文献和资料3.15-3.213学习相关参考文献和资料。3.22-3.284理清思路，撰写开题报告3.29-4.045开题答辩，对设计课题的内容作初步论证4.05-4.116内容论证，内容改进，内容定稿4.12-4.187对小波变换进行研究学习4.19-4.258对小波变换进行研究学习4.26-5.029对各种图像压缩方法研究学习5.03-5.0910比较各种图像压缩方法的优缺点5.10-5.1611熟悉毕业论文格式、撰写论文初稿5.17-5.2312撰写论文初稿5.24-5.3013完成论文初稿并提交5.31-6.0614自我修改毕业论文6.07-6.1315自我修改毕业论文6.14-6.2016根据老师指导，改善不足之处，总体完善6.21-6.2717完成论文终稿，提交论文终稿6.28-7.0418准备好自述讲稿，打印，参加论文答辩四、主要参考文献:1 阮秋琦. 数字图像处理学M. 北京: 电子工业出版社, 20012 赵荣椿. 数字图像处理导论M. 西安: 西北工业大学出版社, 19993 K.R.Castleman. 数字图像处理M. 北京: 电子工业出版社, 20024 钟诚.小波变换及其应用研究J.中国科技信息,2008,25 彭玉华.小波变换与工程应用M.北京:科学出版社,19996 余松煌，张文军.孙军等著.现代图像信息压缩技术M .科学出版社，19987 姜丹.信息论与编码M.合肥中国科学技术大学出版社，20018 黄贤武.数字图像处理与压缩编码技术M .成都:电子科技大学出版社，2009 刘榴梯，刘明奇.党长民等著.实用数字图像处理M .北京理工大学出社，1998.10 马平.数字图像处理和压缩M.电子工业出版社,200711 尹显东，李在铭，姚军等著.图像压缩标准研究的发展与前景M.中国工程物研究院信息与电子工程，2003.12.12 张太怡，吴晓芸，张双腾等著.基于JPEG国际标准的图像压缩方法的研究J .重庆大学学报，1994，17(5)13 AnalogDeviees.ADV-JP2000JPEG2000Co-ProeessorJ.PreliminaryTeehniealData，2001.14 TAYLOR W F. The Geometry of Computer GraphicsM. Wadsworth Inc, 199215 A.J.Patti, M.I.Sezan A.M.Tekalp. Superresolution Video Reconstruction with Arbitrary Sampling Lattices and Nonzero Aperture Time. IEEE Trans, 199716 Elad M Feuer A. Restoration of a Single Superresolution Image from Several Blurred Noisy and Undersampled Measured Images. IEEE Trans, 199717 J.K.Paik, A.K.Katsaggelos. Iamge restoration using a modified Hopfield network. J IEEE Transactions on Image Processing, 1992 18 闫阳,张正炳.基于小波变换的图像压缩编码J.现代电子技术,2005，28(3)19 王晓辉,朱光喜,朱耀庭.图像的一种分形特性表示法及其应用J.电子学报,1997,25(l0)20 练华,宋宝瑞.基于小波变换的分形图像编码J.上海交通大学学报,2004,(4)21 张兢,路彦和.具有感兴趣区域的静止图像压缩编码算法研究J.计算机应用,2005,(6)22 JOEL AM,Larsson C,Charilaos C.Region of Interest Coding in JPEG 2000J.Signal rocessing:Imag Communication,2002,(17)附： 一篇引用的外文文献Analysis of Compression Encoding about Digital ImageYufang Gao Yang LiuBeijing University of Posts and Telecommunications,Beijing,PRC,100876Abstract :This paper mainly investigate the theory of image pressing encoding,two parts,the technology and process of encoding,are included.In the technology of encoding,we give a deep study in thetraditional methods and the use of those methods in the image compression,for example estimation coding,DCT transform,data quantization,entropy coding.Estimation coding lowers the time relativity of image data.DCT transform lowers the space relativity of image data.Data quantization makes use of the redundancy of mentality and vision.Entropy coding brings down the redundancy of coding.After those process,image data will be effectively compressed.Based on the coding technology,this paper examplified MPEG-2 standard discusses the process of the serial of motion image,which includes frame and field coding mode.This paper particularly expounds the field coding of motion compensation combined with DCT transform,introduces the theory of motion compensation and the count of motion vector.Motion compensation makes better use of relativity of field data and increases the compensation ratio.Key words: estimation coding DCT transform motion compensation motion estimation MPEG-21. IntroductionDigital images have many advantages, but the digital image of the massive amount of data hinders the development of digital image technology. In recent years, image compression coding have made rapid development,the sign is International research on image compression coding drawn up a series of compression standards, such as: JPEG, H.26X series, MPEG series。At the same time the development of VLSI technology makes high-performance image coding special chips as possible, which led to the golden age of digital image communication development.Digital color television image signal commonly used component encoding ,luminance signal sampling frequency is 13.5MHZ, color difference signals of the sampling frequency 6.75MHZ. After component encoded, The rate of which composed of three component signals TDM stream is: （13.5+6.75+6.75）8=216Mb/s，so the space for digital video Per minute is: 216Mb/s60s/8=1620MB。Such a huge amount of data makes a 650MB CD can only store one minute of video images, even a 10GB hard drive also can not store a few minutes of video images, it is necessary to compress image data.In this paper,we investigate the method that using MATLAB Wavelet Transform for compression. Methods: The image was processed by wavelet transform, and then the information of low frenquency was retained and the information of high frenquency was home zero. Results: The quality of the first image compression was high, but the compression ratio was low. The compression ratio of the second time was higher,and the image quality was good. Conclusion: The method is simple, reliable and effective in image compression.With the computer technology and the rapid development of network technology, images, sound and other multimedia information recording, storage, transmission has been digitized, the huge amount of data brought to the storage and transmission of certain difficulties, digital image compression has become a solution to the problem key technology. Over the last decade ,wavelet theory has become a new direction in applied mathematics. As a mathematical tool, wavelet was quickly applied to the image and voice of many other fields. Wavelet transform is a signal of time - scale analysis method, the characteristics of multi-resolution analysis, and in time and frequency domain has the capacity of local signal characteristics, a fixed window size, its shape variable, time window and frequency window can change the time-frequency localization analysis, that low frequency high frequency resolution and time resolution, the high frequency part of the high time resolution and lower frequency resolution very suitable for detection of transient entrainment of the normal signal anomalies and show its components, so the microscope known as signal analysis.2, Image Compression Coding TheoryCoding process can be summarized as： Original image Mapping transformation Quantizer Entropy Coder data streamMapping transformation reduces the correlation between image data，Make it more conducive to compression；Quantizer to map the data into a binary digital signal；Entropy code make that appeared on the probability of a large source of symbols assigned to the short code; Small symbol on the probability assigned to a long code, there by reducing redundant data generated code. Data compression is divided into lossless and lossy compression. Lossless compression means that image data after compression can be completely restored, restored the same image and original image, while lossless compression refers to the compressed image data while maintaining the characteristics of the original image under the premise of inevitable loss of part of the original image is not important information. The current image based on wavelet transform compress gradually replaced the DCT and the other based on the coding technique, a new image compression international Biaozhun of choice, such as the most advanced image compression standard JPG2000 the core algorithm is wavelet transform.2.1 Wavelet Analysis Principles and MethodsWavelet analysis is currently applied mathematics and engineering disciplines in a fast-growing new field. Wavelet is a small wave, small means it has the Decay; wave refers to its volatility. With the increasing maturity of wavelet theory, wavelet analysis has become very broad applications. Image processing is an important area of application of wavelet analysis, image processing has become a useful tool.2.2The principle of wavelet transform image compression Wavelet transform for image compression the basic idea is: multi-resolution decomposition of the image, broken down into different space, different frequency sub-image, and then pairs the image coefficients are encoded. Wavelet transform coefficient coding is the core for image compression, compression is the real coefficient quantization compression. Image after wavelet transform wavelet image data generated and the total amount of data equal to the original image, that wavelet transform itself does not have compression. The reason why it used for image compression, because the resulting wavelet image and the original image with different characteristics, and in the image of the energy concentrated in the low frequency part, while the horizontal, vertical and diagonal part of the energy is less; level,vertical anddiagonal part of the characterization of the original image in the horizontal, vertical and diagonal part of the edge information, with clear direction characteristics. Called low-frequency part of the image brightness, horizontal, vertical and diagonal part of the known details of the image Therefore, the simplest method is to use wavelet compression to preserve low-frequency part of the high frequency part of the set 0. Original image is first decomposed into low frequency and high frequency horizontal component L HL1, high-frequency vertical component LH1, high-frequency diagonal components HH1, and then further decompose the low frequency component L repeatedly to the desired level of decomposition. As the scaling function with low-pass filter (H) the role of the wavelet function has a high pass filter (G), the role of image wavelet decomposition for the equivalent of the horizontal and vertical filtering and sub-sampling, the reverse process of the image shall reconstruction 3. 2.3 Image Compression using Mat lab programming stepsCompressed image using wavelet transform under the following three steps: The discrete wavelet transform image is decomposed into low frequency and high frequency components similar to the level of high-frequency vertical and diagonal details of the high frequency component; Extract low frequency, high-frequency part of the set 0; Using the inverse wavelet transform of the reconstructed image .3 ResultsIn this study,we use MATLAB 6.5 programming. Using wavelet decomposition to remove the high frequency part of the image while only retaining the low frequency part of the image as a simple compression method. Function that is used with wavedec2 bior3.7 wavelet decomposition of the image layer, then in low frequency coefficients appcoef2 function, and finally quantified with wcodemat function code. The results are as follows: Original image The first compressed image The second compressed imageImage size before compressionName Size Bytes ClassX 256 256 524288 double arrayGrand total is 65536 elements using 524288 bytesThe first compressed image sizeName Size Bytes Classca1 128 128 131072 double arrayGrand total is 16384 elements using 131072 bytesAfter the second compressed image sizeName Size Bytes Classca2 64 64 32768 double arrayGrand total is 4096 elements using 32768 bytesFrom the above experimental results can be seen that the first compression ratio is 27.81% of before compression. The second compression ratio is 8.58% of before compression. Can be seen from the figure that the effect of the two compression is good, especially the first time the compression is better. The second compressed image darker ,its reason is the loss caused by a large number of low frequency coefficients. It also verified the second compressed greater.4 DiscussionThese results indicate that the image wavelet transform, remove the high frequency part, to retain low frequency, can be better compressed image data, to a certain compression ratio can guarantee the quality of image processing. In addition, to further improve the image compression, but also integrated the use of a variety of other technologies, in particular data encoding and decoding algorithms 4.译文： 数字图像信号压缩的编码分析高玉芳，刘洋北京邮电大学电信工程学院多媒体中心(100876)摘要：本文主要研究图像压缩编码理论，包括两个部分内容：编码技术和编码流程。在编码技术中，我们详细介绍了预测编码、DCT变换编码、数据量化、熵编码的原理及其在图像压缩中的应用。预测编码减弱了图像数据的时间相关性；DCT变换编码减弱了数据的空间相关性；数据量化利用了人眼心理视觉冗余；熵编码减小了编码冗余。减小了这些冗余，图像数据将得到有效的压缩。在编码技术基础之上，本文以MPEG-2为例，系统地讨论了运动图像视频序列的编码流程，包括帧内编码模式和帧间编码模式。其中重点研究了运动补偿帧间预测结合DCT变换的帧间压缩编码，分析了运动补偿的原理和运动矢量的计算方法。关键词：预测编码 DCT 运动补偿 运动估值 MPEG-21 引言图像的数字化有许多优点，但数字图像的海量数据量阻碍了数字图像技术的发展。近年来，图像压缩编码研究取得了飞速发展，其标志国际上图像压缩编码研究成果而制定的一系列压缩标准，如：JPEG、H.26X系列、MPEG系列。同时超大规模集成电路工艺的发展，使得高性能的图像编码专用芯片成为可能，从而引来了数字图像通信发展的黄金时代。数字彩色电视图像信号一般采用分量编码1方式，亮度信号取样频率为13.5MHZ，色差信号的取样频率为6.75MHZ。分量编码后三个分量信号组成的时分复用码流速率为：（13.5+6.75+6.75）8=216Mb/s，则每分钟数字视频所占用的空间为：216Mb/s60s/8=1620MB。这么庞大的数据使得一张650MB的光盘只能存储一分钟的视频图像，即使一块10GB硬盘也存储不了几分钟的视频图像，因此必须对图像数据进行压缩。本文探讨利用MATLAB的小波变换进行图像压缩的方法。方法：对图像进行小波变换，保留低频部分，高频部分置0。结果：第一次压缩时压缩效果较好，压缩比较小，第二次压缩时压缩较大，图像视觉效果也令人满意。结论：本方法是一种简单、有效的压缩方法。随着计算机技术和网络技术的迅速发展，图像、声音等多媒体信息的记录、存储、传输已经数字化，庞大的数据量给存储和传输带来了一定的困难，数字图像的压缩已成为解决该问题的关键技术。近十几年来小波理论的研究己成为应用数学的一个新方向。作为数学工具，小波被迅速应用到图像和语音分析等众多领域。小波变换是一种信号的时间尺度分析方法，具有多分辨率分析的特点，而且在时频两域都具有表征信号局部特征的能力，是一种窗口大小固定不变但其形状可变，时间窗和频率窗都可变的时频局部化分析方法，即在低频部分具有较高的频率分辨率和时间分辨率，在高频部分具有较高的时间分辨率和较低的频率分辨率，很适合探测正常信号中夹带的瞬态反常现象并展示其成分，所以被誉为信号分析的显微镜2 图像压缩编码原理图像编码的过程可以概括为：原始图像 映射变换 量化器 熵编码器 码流映射变换减小了图像数据之间的相关性，使之更有利于压缩编码；量化器将映