基于嵌入式思想的小波图像压缩研究

基于嵌入式思想的小波图像压缩研究

摘要:图像作为传递信息的媒体和手段是十分重要的。但是数字图像的数据量非常大,需要很多的比特数,因而需要对图像进行压缩。所谓图像压缩就是在保证一定的图像质量和满足一定要求的前提下,减少图像原始数据量的一种过程。
关键字:图像压缩嵌入式小波变换
Abstract:
The image as delivering medium and means of information are very important the data quantity of the numerical image is very big, needing a lot of ratios to especially count, as a result need to carry on the compression to the image -called image compression is at guarantee the certain image quantity and satisfy the premise of certain request under, a process of reducing the quantity of original data of image.
Firstly, the image compression's necessary and feasibility are discussed, and several classical image coding technologies are introduced. Including code method and its evaluation criteria, main image coding international standard and so on. Finally, the Embedded Zerotree Wavelet algorithm based on the wavelet transformation had been analysised and the discuss, and its basic thought、 concept and algorithm principle had been elaborated. And the programming based on matlab has been completed: overall situation threshold value image compression Based on the bior3.1 wavelet, image two-times compressions and the primitive image 3-times decompositions based on the bior3.7 wavelet, the image compression-encoding and decoding based on the harr wavelet and the huffman coding. Through carries on the compression code test experiment to the standard test image, confirmed this method operation’s succinction, the quick speed, the high quality of the re-build images, which has obtained the satisfying effect.
Keywords:Image compression Embedded Wavelet transform
本文首先论述了图像数据压缩的必要性和可行性,对传统的经典的图像编码技术进行了介绍。包括编码方法及其评价标准、主要的图像编码国际标准等。并对基于小波变换的嵌入式零树算法进行了详细的分析和讨论,阐述了嵌入式零树小波编码(EZW)的基本思想、概念及算法原理。并完成了基于harr小波基和huffman编码方法的图像压缩编解码的编程。通过对标准测试图像进行压缩编码测试实验,证实该方法运算简洁,速度快,重构图像质量高,取得了令人满意的效果。
一、图像压缩编码概述
1.图像压缩编码的`原理
图像压缩机制通常可以分为两种:有损压缩和无损压缩。在无损压缩中,我们关心的是确信重建没有丢失信息,无损压缩通常被用于文本文件的压缩中。对于有损压缩,允许压缩后产生质量上的误差。有损压缩机制的优点是可以得到比无损压缩高得多的压缩比,但是它只能用于可以用近似的数据代替原始数据,而这种相近数据又是容易被压缩的情况。举例来说,一幅和原图像在视觉上没有区别的图像可以看作是原图像的近似。任何压缩机制的根本思想都是除去数据中存在的相关性。所谓相关性,就是能够根据给出的一部分数据判断其相邻的数据。数据相关性分为以下几种:
(1)空间相关性:可以根据图像中某一点的像素值推断出其相邻点的像素值;
(2)频率相关性:一个信号的傅立叶变换通常是光滑的,这意味着可以根据某一部分的频率来推断其相邻部分的频率;
(3)时间相关性:在数字视频中,在时间上相邻两帧图像的大部分像素的值变化很小。
有损压缩的标准过程是变换编码。基本思想是用一个和原来不同的数学基来表示数据,在这种新的表示下,数据的相关性能够显露出来或被拆开。在这种情况下,大部分的系数都接近于0,可以忽略,于是可以将余下的信息存储在一个较小的数据包中。