静态图像压缩方法研究

【作者】 伞兴；

【导师】 吴秀清；

【作者基本信息】 中国科学技术大学 ， 信号与信息处理， 2007， 博士

【摘要】 当前社会随着信息技术的飞速发展，人们对图像质量、尺寸、读取速度的要求越来越高，图像压缩已经成为数据压缩的一个核心组成部分，并一直吸引着广大研究人员的注意。目前以国际编码标准JPEG2000为代表的新一代基于小波变换的静态图像压缩方法，在将图像压缩效率提升到一个新的高度的同时，还能提供关于时域、空域、质量等多种可扩展性。但这些辉煌的成就并不能满足人们日益增长的需求，图像压缩技术还需要继续进行深入的研究。本文在已有的图像压缩方法的基础上，分别分析了灰度图像、多通道图像以及多视角图像的统计特性，并提出了几种新颖的、更有效的图像编码算法。本文的主要工作和创新之处归纳为以下几点：1．从条件信息熵的角度出发，在理论上分析了高维上下文模型的最优量化标准，并提出了一种具有较低计算复杂度的上下文量化标准。2．结合离散小波变换和基于上下文的自适应算术编码方法，提出了一种基于时、频域上下文预测模型的图像压缩算法。该算法首先建立了一个时、频域的上下文预测模型，然后合理地对该模型中的上下文进行量化，以得到合适的编码上下文用于自适应的算术编码以降低模型开销。3．通过对以下两个问题的解答：1)彩色图像中亮度和色度分量之间到底存在什么性质的相关性?2)这种相关性到底有多强烈?找出了最适合进行颜色编码的上下文，并进一步地提出了一种新的嵌入式彩色图像编码方法。4．提出了一种基于三维小波变换和上下文量化的高光谱图像压缩算法。该算法利用三维小波变换去除高光谱图像在谱内和谱间的冗余信息，然后建立了一个高维时、频域的上下文预测模型以去除小波系数之间的冗余信息。5．在多视角图像压缩这一新兴领域，提出了一种能够获得准确差异向量(DV)的多视角几何预测方法。在这一预测方法的基础上，进一步地设计了一个通用的多视角图像编码框架。综上所述，本文对静态图像压缩技术进行了深入的研究，取得了一些有价值的研究成果。数字图像压缩仍然是一个较有潜力的研究领域，值得我们更深入的研究。更多还原

【Abstract】 With the fast development of information technology, the requirements on image quality, size, and transmission speed presently become more and more intense. As one of the kernel components in data compression field, image compression always attracts many researchers’ attentions. Now, the new generation of image coding method, such as JPEG2000 standard, has put the efficiency of image compression on a very high level. At the same time, these methods also provide some useful functions: temporal scalability, spatial scalability and quality scalability. However, these splendent achievements still can not satisfy the people’s requirement. There are a lot of works still should be done to improve the performance of image coding.Based on the existent research works, this paper respectively analyses the features of gray-scale image, multi-channel image and multi-view image. Based on the clear analysis, several novel and efficient image coding methods are proposed. The main contribution of this dissertation can be summarized as follows:1. We studied the optimal context quantization criterion for the high dimension context model from a view of conditional information entropy. Then, a novel context quantization criterion with low computational complexity is proposed.2. The paper describes an image coding algorithm based on the discrete wavelet transform and context-based adaptive arithmetic coding. A novel coding model, the combination of the spatial and frequency prediction in the wavelet domain, is proposed in the paper. At the same time, context quantization as the key part of arithmetic coding is carefully analyzed in order to obtain in suitable contexts for coding and decrease the model cost.3. Analyzes the inter-color correlation and answer two questions related to color image coding: (1) what kind of inter-color correlation exists in color images after the discrete wavelet transform? (2) How strong is it? This analysis helps us to find a most suitable inter-color context and eventually leads to a new embedded color image codec. By using the discovered inter-color context, significant performance improvement can be achieved when encoding chrominance components.4. Proposed a novel hyper-spectral image coding method based on three-dimension wavelet transform (3DWT) and context quantization. In this method, 3DWT is firstly adopted to remove the intra- and inter-spectral redundancy of hyper-spectral images. Then the correlation of wavelet coefficients is analyzed bya high dimension context predicting model. 5. In the booming field of multi-view coding, proposes a geometric predictionmethodology for accurate disparity vector (DV) predicting. Based on the new DVpredictor, this paper designs a basic framework that can be implemented in mostexisting multi-view image/video coding schemes.In conclusion, we studied the static image coding techniques and achieved some valuable results. However, static image coding is still a potential research field which is worthy of further study.更多还原