基于directionlets变换的EZW图像压缩编码算法分析-模式识别与智能系统专业论文.docxVIP

摘要 摘要 近些年来，随着因特网在全球的日益普及和移动通信的迅猛发展，以及新兴 多媒体业务的出现，对图像/视频技术压缩编码技术提出了许多更高的要求。理 想的图像压缩编码技术除了提供传统的良好压缩性能与重建质量外，还应该提供 关于质量、分辨率、信噪比等多种可扩展编码结构，并且实现多分辨率编码、嵌 入式编码及抗误码传输，从而能在无线移动环境向用户提供个性化的服务。 小波变换理论可以用在图像处理领域，但是效果不是十分的理想，原因是标 准小波变换的变换基只有垂直和水平两个方向，对图像的边缘或轮廓等特征信息 的表达不是最优的稀疏表达。 为了更好的刻画图像的几何特征信息，如边缘或轮廓，本文全面系统地研究 了 directionlets 变换的理论，并对其在图像压缩领域中的应用作了研究，主要工 作在以下两个方面： (1) 基于 directionlets 变换对图像的分解与重建。深入研究了 directionlets 变 换的变换基构造，并为此研究了构造可分离滤波器组的方法。 (2) EZW 图像压缩编码算法充分利用了标准小波变换的空间多分辨率特征， 通过对重要节点以及零树根编码达到压缩目的，将 directionlets 和 EZW 结合起来， 应用零树编码方案进行图像编码。 实验结果表明：相对于传统的基于标准小波变换的 EZW 编码，基于 directionlets 变换的 EZW 算法在峰值信噪比以及主观视觉效果方面都更为优越。 关键字：小波变换；方向小波变换；嵌入式零树小波编码；图像压缩 i Abstract Abstract Recently, with the increasing popularity of the Internet in the world and the rapid development of mobile communications, as well as the emergence of new multimedia services, image / video compression technology has been made many higher demand. Ideal image compression coding technology in addition provide traditional good compression performance and reconstruction quality, it should also provide information on quality, resolution, signal to noise ratio and other scalable coding structure, and multi-resolution coding, embedded coding and anti- transmission error, which can provide users with wireless mobile environment personalized service. Wavelet theory can be used in image processing, but the effect is not very ideal because standard wavelet transform(WT) based on only vertically and horizontally directions, it fails to provide a optimal sparse representation of feature information,such as edges and contours of a smooth image. In order to obtain better geometric feature information, such as edges or contours, this paper systematic study of the directionlets transform theory, and its application in the field of image compression, mainly in the following two aspects: Effective representation method of smooth image using directionlets transform. We further studied the transformation structure of direction