【作者】 宋玉霜；

【导师】 王磊；

【作者基本信息】 西安理工大学 ， 计算机应用技术， 2010， 硕士

【摘要】 针对互联网个人影像产品制作过程中,传统手工操作复杂、耗时、并造成网络与服务器高负荷运转的问题,探讨了一种基于等级变异与图像内容特征信息的免疫克隆算法。并以比较复杂和典型的个人影像产品照片书为例,讨论其个性化制作方法。该方法首先根据制作照片书过程中所涉及的一些因素之间的相互关系,特别是色彩学、色彩搭配学、色彩的心理感受等艺术专业知识,以及相关领域专家们的建议提出了8个评价标准来衡量一本照片书的质量。这些标准包括：(1)照片书封皮与其背景的融合度；(2)照片书造价与底价的贴合度；(3)每页背景与主色调的融合度；(4)每页背景与该页照片的融合度；(5)每页背景与该页照片的对比度；(6)每页背景与该页照片的搭配度；(7)每页字色与该页背景的融合度；(8)每页字色与该页背景的对比度。其次,论文分析了将颜色特征作为图像主要内容提取过程中所涉及的一些技术问题。特别是在量化的HSV空间内,探讨了针对本问题的颜色区间概念、图像模型及其计算方法。进而在图像距离和颜色区间距离计算的基础上,给出了上述8个指标的实施方法,实现了指标的量化。最后,根据本问题的各种特性,探讨了一种专门针对个人影像产品处理的基于等级变异的克隆选择算法来优化其制作过程,实现影像产品抗体群的快速收敛与目标选定。理论分析与仿真结果表明,该方法应用于目前互联网的工作环境,具有良好的可行性、实用性和有效性。目前,该算法已经被采纳并实现于我们开发的照片产品制作网站PanoPhoto中,系统的运行状况以及客户的反馈信息支持了算法的仿真结果。更多还原

【Abstract】 For the problems in the process of making personal photo products with traditional manual based operations, such as complicated procedures, time-consuming and high loaded process of the network or servers. A novel individual image processing oriented immune algorithm is proposed, which utilizes some characteristic information such as a grade-based mutation and content image. And the paper takes photo books which are more complex, typical personal photo products as example to discuss its personalized making.In this method, according to relationship of some factors which are involved in the process of making photo books, in particular, chromatics, color collocational theory, color psychological, etc, and suggestions offered by experts in related fields, we proposal 8 evaluation criterions to measure the quality of photo books. These include:(1) Degrees of fusion between photo book cover and background; (2) Degree of closeness between the practical cost and reserve price; (3) Degrees of Fusion between Every Background and Main Color; (4) Degrees of fusion between per page’s background and photos; (5) Contrast between per page’s background and photos; (6) Degrees of collocation between per page’s background and photos; (7) Degrees of fusion between per page’s font color and background; (8) Contrast between per page’s font color and background.Secondly, we extract color feature as the main content of photos. And in quantized HSV color space, the concept of color interval, the model of image, and their methods of calculation are proposed. Then based on the calculation of image distance and color interval distance, the paper proposes methods of calculation for the above-mentioned 8 evaluation criterions to achieve their quantification.Finally, for the problem of making personal photo products, according to the various features of this issue, we propose a grade-based mutation clonal selection algorithm to optimize their making process, and achieve fast convergence of antibody population and the target selected.Theoretical analysis and simulation results show that the method is applied to the current working environment of the Internet and has good feasibility, practicality and effectiveness. Currently, the algorithm has been adopted and implemented in our photo production making website PanoPhoto, the system’s operating conditions and customer feedback support the simulation results of algorithm.更多还原