【作者】 黄鹏涛；

【导师】 陈贤富；

【作者基本信息】 中国科学技术大学 ， 电路与系统， 2010， 硕士

【摘要】 在保证图像质量的条件下,对图像进行压缩处理,实现高效的图像存储、传输是人们一直研究的热点问题。传统的压缩方法已经很难更大幅度的提高压缩率,人们开始对图像压缩技术引入新的理论和方法,探索新的压缩算法。随着人工智能等相关技术的发展,各种智能计算方法在图像压缩方向的应用不断出现,探索智能计算在图像压缩中的应用成为了一个新的研究方向。元胞自动机(CA)是人工生命的重要研究工具和理论方法分支,CA理论在图像压缩领域的应用研究也是一个新的研究热点。本文利用CA演化状态非线性、多样性、并行性等优点,对基于CA模型的图像压缩算法进行了研究,主要研究工作和创新之处有以下几点。本文首先简要概述了几种典型的元胞自动机模型的基本理论和发展状况,并对其在图像压缩领域的应用做了相应的理论分析,根据不同模型特点提出了其在图像压缩方向的优势和不足。其次,根据CA模型自组织行为,变换状态的多样性的并行处理等特点,提出了基于CA模型的二值图像压缩算法的并行压缩方法;不同于传统的基于去相关、去冗余压缩方向的局限性,该压缩方法用遗传算法搜索最优的元胞自动机规则,利用矢量量化的思想,但避开了码本的设计和搜索复杂度高的缺点,达到了类似的压缩效果,实现了更高效的快速编码算法。第三,研究了基于CA模型的灰度图像矢量量化压缩编码算法,利用多状态CA的并行处理特性,实现快速的码字搜索算法,使搜索时间复杂度大大低于传统的码字搜索算法。第四,将模糊逻辑引入到元胞自动机模型,探索基于模糊元胞自动机模型的矢量量化中码本训练算法,利用模糊信息的不确定性,并行特点,优化生成的码本,使元胞自动机模型的应用能力更好,更智能,丰富了元胞自动机的理论。第五,研究了基于元胞神经网络(CNN)模型在图像压缩方面的应用,提出了用CNN模型实现DCT算法,极大的提高了压缩算法效率,并分析了其算法优越性。最后对论文的研究内容进行了总结,并提出了基于CA模型在图像压缩领域进一步的研究工作。更多还原

【Abstract】 Image compression becomes increasingly important to support efficient storage, transmission, but the traditional compression algorithms can no longer meet the need of achieving much higher compression. Researchers in the field of image compression have focused their attention on the study of some new kinds of compression model such as artificial life. Theory and applications of the Cellular Automata (CA) as one of the directions of artificial intelligence has become an important research field of data compression in recent years.CA is dynamical systems exhibiting many notable features, namely, nonlinear, massive parallelism, and discreteness locality of cellular interactions, etc. In this paper,we study Cellular Automata (CA) as a modeling tool successfully using in image compression. The dissertation is organized as follows:First of all, the theoretic and analysis of some CA modeling are reported in Chapter2. We describe the advantages and disadvantages of the efficient compression algorithm basing on CA modeling compared with the traditional methods. Their potential within image compression is being investigated.In Chapter3 reports a new Cellular Automata (CA) model for binary image compression. The search for appropriate CA having the desired characteristics to system is an extremely difficult task in view of exponentially large search space, The genetic programming (GP) has been employed to search for optimal non-linear cellular automata rules which performs the binary image compression task. The simulation research proves that the algorithm is feasible and more efficient in compression ratio, compression speed, decompression precision and the code can be compressed using other compression algorithms etc.Compression of grayscale image data using CA model is also proposed. We presents a new quick code scheme of vector quantization based on CA. The effective- ness of code was evaluated by comparing the other methods.Chapter4 reports the characterization of fuzzy cellular automata. The analytical study the possible compression using Fuzzy cellular automata has been done and the Algorithm is also proposed and analyzed.Chapter5 describes the image compression method based on Cellular Neural Networks (CNN). The parallel CNN model is used for DCT image data compression. Finally Chapter 6 concludes the volume, presenting several possible avenues of future research.更多还原