【作者】 王善侠；

【导师】 王英健；

【作者基本信息】 长沙理工大学 ， 计算机应用技术， 2005， 硕士

【摘要】 随着多媒体技术和计算机通信的日益发展,具有庞大数据量的数字图像极大地制约了图像通信。采用有效的压缩编码技术删除冗余,以尽量少的比特数存储图像,并同时保证图像的质量,己成为研究的热点。分形图像可进行压缩的原因是其图像具有高度的自相似仿射特性。本论文从理论和实践上介绍了不同于传统方法的压缩算法—基于遗传算法与视觉特性的分形图像压缩算法。作者采用迭代函数系统(IFS)对图像进行压缩。主要以不动点定理和拼贴定理作为理论基础,对给定的图像,寻找一组由压缩仿射变换构成的IFS,使图像通过仿射变换后尽可能与其相似。图像的解码,不依赖于原始图形,对任意初始图像,用IFS 反复迭代,就能将原始图像重现。因此,编码文件只需存储IFS 码,从而能得到较高的压缩比。　分形图像压缩的最终目的是要获得一个较好的IFS,使它的吸引子与原始图像尽可能的相似,可以把这个搜索问题看成是一个优化过程。这个优化过程是在大空间搜索和具有许多复杂约束的知识背景下进行的。传统的经典算法难以解决此问题,于是作者采用具有人工智能技术的遗传算法来寻找最优解。遗传算法从种群开始搜索,每个种群所获得的知识都被嵌入了其成员的染色体中。在进化过程中,它引入了自然进化和适者生存的原理,在复杂和变化的环境中寻找最有利的生存方式,可以以较大的概率较快速地找到最优解。　论文首先介绍了迭代函数系统和遗传算法的基本定理和原理,它们是该课题研究的理论基础。然后,介绍了传统的分形算法。最后作者提出了一种改进的用于求解具有全局最优的自相似分块匹配的遗传算法,详细地一讲述了染色体编码方法、适应度函数的设计、遗传算子的设计和采样机理。并且利用人眼对灰度变化的敏感程度与背景有关的特点,在综合考虑图像块的均值与方差的前提下,动态地改变分形压缩编码时定义域块仿射变换后拼贴到值域块的最小误差,从而提高了编码速度。通过实验的结果表明,该方法解码质量好、编码速度快,是一种可行的算法。更多还原

【Abstract】 With the developing of multimedia technology and computer communication, digital image which have enormous data quantity restricts image communication. Effective encoding technologies to obliterate redundance and retain the image quality are the focus of research. The possibility of compression is because of high self-similarity and self-transformability redundancy of images. In this dissertation, from theoretical and practice viewpoint, we present a novel approach to compress gray-scale images based on Genetic Algorithms and Visual Character. The kernel theory based on in this paper is called Iterated Function System (IFS). The main idea is to find an IFS which consists of a set of contractive affine transformations mainly based on fixed-point theorem and collage theorem, when they are applied on the original image, the union of the transformed images will cover up the original image. Decoding process starts from any images which can recur the original image by applying IFS. Therefore, coding file only stores IFS code, which can achieve high compression ratio. The purpose of fractal image compression is to gain a good IFS whose attractor is similar to original image. So this search problem can be viewed as a combination and optimization process with complicated constraints and a lame searching space. Because traditional algorithm hardly handle with this problem, we adopt Genetic algorithm to solve this problem which has the property of artificial intelligence technology. Genetic algorithm starts from initial population. The individuals of each population embed some new information while the searching can be directed to the promising area. The evolutive principle and the survival mechanism are applied in the algorithm. As a result, the best result can be achieved with high probability. The dissertation starts by explaining the basic notions of Iterative Function System and Genetic Algorithms. Then we go to details of the ideas of compression algorithms based on IFS theory, Lastly, an Genetic algorithm is proposed for obtainment of matching domain blocks of fractal partition in image compression. It makes use of the partition iterated function system and fractal image compression. Chromosome representation, initialization of population, design of special genetic operators are introduced explicitly. In this coding algorithm,based on the character that the visual sensitivity of eye relates to the background,we dynamically modify the matching error after sufficient considering the average and the difference of the coding block,consequently reduce the coding time.The algorithm is robust and optimal. Both theoretical analyses and experiments show that higher image quality can be achieved. Parallel computation of genetic algorithms could reduce time cost in fractal compression.更多还原