近似最优最近邻高维局部线性扫描算法研究-计算机技术专业论文.docxVIP

Abstract An important research issue in the field of image databases is the content-based image retrieval．High-dimensional indexing has been considered all important means to improve the performance in querying large image databases，and it has been a very active research area in multimedia and database applications over the last few years．It has been observed that the performance of traditional indexing methods deteriorate rapidly as dimensionality increases-a phenomenon which has become known as the ‘curse of dimensionality’ .Efficient indexing schemes for high-dimensional data are required for real-time querying in large-scale image database ． High-dimensional indexing also can be used in other research area ， such as data mining ，pattern recognition，machine learning and data analysis． The phenomenon of ‘curse of dimensionality’ is introduced in this dissertation firstly,and the current status and future development trend of high-dimensional indexing are summarized.For the efficiency of query in high-dimensional spac,firstly two typical kinds of one-dimensional mapping method, direct projection and distance computation regarding to reference points, are discussed in this dissertation. An optimal combination of one-dimensional mappings is achieved for the best search performance. Furthermore, we propose a near-optimal partial linear scan algorithm by considering several one-dimensional mapping values. Secondly, The vector approximation approach is an efficient high-dimensional indexing method,whose performance still outperforms sequential scan in high-dimensional space. A new high-dimensional indexing structure based vector approximation method and One-dimensional mapping method was introduced.Only a small set of approximate vectors are accessed during the search， which can reduce the computational complexity and i/o cost.Lastly a new indexing structure based on clustering with Gaussian Mixture is also designed to facilitate the partial linear scan