QQ客服 官方微博 反馈留言

题名	DCT域内数字图像特征自动提取的方法与应用研究
作者	李鹏杰
学位类别	博士
答辩日期	2004
授予单位	中国科学院声学研究所
授予地点	中国科学院声学研究所
关键词	特征提取 图像检索
其他题名	Theory and Implement Study on Image Feature Automatic Extraction in DCT Domain
中文摘要	随着图像处理技术研究的深入和多媒体应用的发展，图像处理领域的相关研究已经从图像效果处理、图像视频编码转移到智能化的高级应用方面，比如对象的提取、识别，基于内容的图像检索，目标跟踪等等。特征自动提取作为这些应用的基础和核心问题也越来越受到关注，同时一由于我们继承了互联网和通讯技术发展带来的后效应—大量的压缩后的数据流，使得在压缩域进行图像特征提取也成为近年来的研究热点。而在这些压缩后的码流中，绝大部分是使用DCT的方法进行压缩的，本论文的研究目的就是想获得高效的DCT域内图像自动特征提取的方法，这样可以避免运算量巨大的解压缩过程进而提高处理速度。DCT域内图像特征提取的研究在国际上已经开展了一段时间，取得了不少成果，本论文在这些研究成果的基础上进行了新的创新性工作。在理论上，我提出了一种新的基于线性代数的DCT域内块分割方法，这种方法可以把任意大小的矩形DCT块在频域内直接分解成多个子块，子块也可以是任意大小的矩形DCT块。它避免了原有基于三角变换的方法带来的复杂的推 导过程，同时把原有方法只能进行2的整数次幂的分解扩充到任意大小。这种方法扩充了DCT域内可进行的操作方式，使在DCT域内进行与像素域相同效果的操作成为可能。在方法上，基于上述的新理论我开发了一种直接在DCT域内提取MPEG7主颜色描述符的方法。这种方法利用DCT系数的统计特征与像素统计特征的映射关系，以DCT域内的块分割为理论基础，在DCT域内实现了像素域内相同效果的操作，提取了图像主颜色构建了MPEG7描述符。同时新算法还能够自动确定闭值，解决了多数闻值需要经验确定的问题。与像素域内构建的MPEG7描述符的对比检索实验表明，DCT域内直接提取的描述符反而具有稍高的准确度。此外论文中还介绍了我基于MPEG2进行的自适应帧分类的研究，以及开发的一种新的基于内容的图像检索系统ImageHunter。
英文摘要	The research of image processing focus on intelligent implement from image enhancement and image compression & coding, while the technique and multimedia application is developing faster and faster. Intelligent implements include object extraction and recognition, contend-based image retrieval, object tracking etc. al. As the basis and core problem of these issues, feature automatic extraction is attracted attention by more and more researchers. Furthermore, the derivation of the development of Internet and communication, quite a large amount of compressed bit streams, make researchers pay more attention to extract image features in compressed domain. Among these compressed bit streams, most of them are compressed based on DCT. This dissertation aims to find an effect method of image feature extraction in DCT domain. It can make image processing faster by avoiding decompression process. DCT domain Image feature extraction has been researched for some years. Some achievements have been got. Based on these work, I did some new creative research and represented in this dissertation. For theory, I found a new method based on linear algebra to split a DCT block in DCT domain. This method can conveniently split an arbitrary rectangle DCT block into some rectangle DCT sub-blocks in compressed domain. The original method based on triangle transform can only split square block into square sub-blocks, and the derivation is very complex. My method supplies a new one for image manipulation in DCT domain. It make people could process image in DCT domain as same as process that in pixel domain. For method, I developed a new method to extract MPEG7 dominant color descriptor in DCT domain based on the theory above. It uses the relationship between statistical property of DCT coefficients and statistical property of pixel. Based on the theory of splitting block in DCT domain, the method realizes the same effect in DCT domain as in pixel domain. It extracts dominant colors of image and construct MPEG7 descriptor. New method can determine thresholds automatically, not detemiined by experience. Experiments of image retrieval compare these two MPEG7 descriptors constructed in DCT domain and pixel domain respectively. The results show that the descriptor extracted in DCT domain is a bit more accurate than the other one. Besides these two contend, the dissertation also represents my research on adaptive classification of MPEG 2 video frames and a new contend based image retrieval system: ImageHunter.
语种	中文
公开日期	2011-05-07
页码	106
内容类型	学位论文
源URL	[http://159.226.59.140/handle/311008/834]
专题	声学研究所_声学所博硕士学位论文_1981-2009博硕士学位论文
推荐引用方式 GB/T 7714	李鹏杰. DCT域内数字图像特征自动提取的方法与应用研究[D]. 中国科学院声学研究所. 中国科学院声学研究所. 2004.

个性服务

查看访问统计

相关权益政策

暂无数据

收藏/分享

除非特别说明，本系统中所有内容都受版权保护，并保留所有权利。