• Journal of KIISE:Databases
• /
• v.29 no.3
• /
• pp.193-206
• /
• 2002

In this paper, we design and implement an efficient bulk-loading algorithm for CIR-Tree. Bulk-loading techniques increase node utilization, improve query performance and reduce index construction time. The CIR-tree has variable size of internal node entries since it only maintains minimal dimensions to decriminate child nodes. This property increases fan-out of internal nodes and improves search performance. Even though several bulk-loading algorithms for multi/high-dimensional index structures have been proposed, we cannot apple them to CIR-tree because of the variable size of internal node entries. In this paper, we propose an efficient bulk- loading algorithm for CIR-tree that improves the existing bulk-loading algorithm and accomodates the property of CIR-tree. We also implement it on a storage system MiDAS-III and show superiority of our algorithm through various experiments.

• The KIPS Transactions:PartD
• /
• v.10D no.1
• /
• pp.1-12
• /
• 2003

Recently, there have been many indexing schemes for multimedia data such as image, video data. But recent database applications, for example data mining and multimedia database, are required to support multi-user environment. In order for indexing schemes to be useful in multi-user environment, a concurrency control algorithm is required to handle it. So we propose a concurrency control algorithm that can be applied to CBF (cell-based filtering method), which uses the signature of the cell for alleviating the dimensional curse problem. In addition, we extend the SHORE storage system of Wisconsin university in order to handle high-dimensional data. This extended SHORE storage system provides conventional storage manager functions, guarantees the integrity of high-dimensional data and is flexible to the large scale of feature vectors for preventing the usage of large main memory. Finally, we implement the web-based image retrieval system by using the extended SHORE storage system. The key feature of this system is platform-independent access to the high-dimensional data as well as functionality of efficient content-based queries. Lastly. We evaluate an average response time of point query, range query and k-nearest query in terms of the number of threads.

• Proceedings of the Korean Information Science Society Conference
• /
• 2000.04b
• /
• pp.134-136
• /
• 2000

본 논문에서는 고차원의 특징 벡터 공간에서의 객체에 대한 효율적인 검색을 지원하는 셀기반 시그니쳐 트리 색인 구조(CS-트리, CI-트리)를 제안한다. 특징 벡터 공간을 셀로써 분할하고 특징 벡터는 셀의 시그니쳐로 표현되며 트리에 저장된다. 특징 벡터 대신 시그니쳐를 사용하여 트리의 깊이가 낮아짐으로서 검색을 효율적으로 수행할 수 있다. 또한 셀에 적합한 새로운 가지치기 거리를 이용한 유사성 검색 알고리즘으로 수행할 수 있다. 또한 셀에 적합한 새로운 가지치기 거리를 유사성 검색 알고리즘을 제시한다. 마지막으로 우수한 고차원 색인 기법으로 알려져 있는 X-트리와 성능 비교를 수행하여, 성능비교 결과 본 논문에서 제안하는 CS-트리와 CI-트리가 검색 시간 측면에서 최대 30%의 검색 성능이 개선됨을 보인다.

• Proceedings of the Korea Information Processing Society Conference
• /
• 2007.11a
• /
• pp.211-214
• /
• 2007

최근 멀티미디어 데이터, 특히 UCC를 중심으로 동영상 데이터가 급증하고 있다. 그러나 현재 대부분의 검색 시스템은 키워드 기반의 동영상 데이터 검색만을 지원하고 있으며, 따라서 사용자가 원하는 동영상 데이터를 효율적으로 검색하지 못하는 실정이다. 동영상 데이터에 대한 효율적인 검색을 지원하기 위해서는, 동영상의 내용(이미지, 색, 모양 등)을 고차원의 특징 벡터 데이터로 표현하여 유사한 동영상을 검색하는 내용-기반 검색이 요구된다. 본 논문에서는 내용-기반 검색을 위해 제안된 기존의 고차원 벡터 데이터 색인 구조를 실험을 통하여 성능을 비교하며, 이를 통해 동영상 내용-기반 검색에 가장 효율적인 색인 기법을 제시한다. 아울러 보다 효율적인 내용-기반 검색을 위한, 근사 k-NN 질의 탐색 기법의 유용성을 검증한다.

• Proceedings of the Korean Information Science Society Conference
• /
• 2000.04b
• /
• pp.163-165
• /
• 2000

내용기반 이미지 검색을 위한 기존의 대부분의 기법들은 이미지 데이터에 효과적으로 적용할 수 있는 고차원의 색인구조를 고려하지 않았다. 이 연구에서는 이미지 데이터베이스에서 보다 효율적이며 정확도가 높은 검색결과를 기대할 수 있는 색상 특징 데이터 표현방법인 ECCV기법, 모양 특징 데이터 표현방법인 EPA기법을 소개한다. 또한 고차원 데이터에 대해서도 검색속도를 향상시킬 수 있는 새로운 다차원 공간 인덱스 구조인 XS-트리를 제안한다. 이 방법을 이용하면 특징표현단계에서는 차원의 수가 증가되어 저장에 필요한 공간을 많이 요구하지만 인덱싱 단계를 거치면 이미지 검색 속도가 향상되며 정확한 이미지를 검색 할 수 있는 장점이 있다.

• The KIPS Transactions:PartD
• /
• v.8D no.4
• /
• pp.305-312
• /
• 2001

Recently, high-dimensional index structures have been required for similarity search in such database applications s multimedia database and data warehousing. In this paper, we propose a new cell-based signature tree, called CS-tree, which supports efficient storage and retrieval on high-dimensional feature vectors. The proposed CS-tree partitions a high-dimensional feature space into a group of cells and represents a feature vector as its corresponding cell signature. By using cell signatures rather than real feature vectors, it is possible to reduce the height of our CS-tree, leading to efficient retrieval performance. In addition, we present a similarity search algorithm for efficiently pruning the search space based on cells. Finally, we compare the performance of our CS-tree with that of the X-tree being considered as an efficient high-dimensional index structure, in terms of insertion time, retrieval time for a k-nearest neighbor query, and storage overhead. It is shown from experimental results that our CS-tree is better on retrieval performance than the X-tree.

• Journal of KIISE:Databases
• /
• v.29 no.1
• /
• pp.58-71
• /
• 2002

Generally, multi-dimensional data such as image and spatial data require large amount of storage space. There is a limit to store and manage those large amount of data in single workstation. If we manage the data on parallel computing environment which is being actively researched these days, we can get highly improved performance. In this paper, we propose a parallel high-dimensional index structure that exploits the parallelism of the parallel computing environment. The proposed index structure is nP(processor)-n$\times$mD(disk) architecture which is the hybrid type of nP-nD and lP-nD. Its node structure increases fan-out and reduces the height of a index tree. Also, A range search algorithm that maximizes I/O parallelism is devised, and it is applied to K-nearest neighbor queries. Through various experiments, it is shown that the proposed method outperforms other parallel index structures.

• The Transactions of the Korea Information Processing Society
• /
• v.7 no.8
• /
• pp.2327-2340
• /
• 2000

Existing bulk loading algorithms for multi-dimensional index structures suffer from satisfying both index construction time and retrieval perfonnancc. In this paper, we propose an efficient bulk loading algorithm to construct high dimensional index structures for large data set that overcomes the problem. Although several bulk loading algorithms have been proposed for this purpose, none of them improve both constnlCtion time and search performance. To improve the construction time, we don't sort whole data set and use bisectiou algorithm that divides the whole data set or a subset into two partitions according to the specific pivot value. Also, we improve the search performance by selecting split positions according to the distribution properties of the data set. We show that the proposed algorithm is superior to existing algorithms in terms of construction time and search perfomlance through various experiments.

• Journal of KIISE:Computing Practices and Letters
• /
• v.7 no.5
• /
• pp.392-401
• /
• 2001

Nowadays, the amount of image data increase explosively. To manage the large amount of image data efficiently, high-dimensional index structures are necessary. However, as my knowledge none of existing DBMSs supports high-dimensional index structures as access methods of DBMSs. In this paper, we design and implement CIR-Tree as a access method for retrieving image data effectively on the MiDAS-III that is the storage subsystem of the BADA-III. The implemented CIR-Tree manager shows much better retrieval performance than sequential search in performance evaluation.

• Proceedings of the Korea Contents Association Conference
• /
• 2019.05a
• /
• pp.443-444
• /
• 2019

효율적인 이미지 검색을 위해 고차원 데이터 색인에 대한 연구가 진행되고 있다. 거리 기반 색인 구조는 다차원 데이터를 색인하는데 자주 활용되는데, k-최근접 질의 처리에서 초기 탐색 범위를 전체 영역의 1%만으로 결정한다. 본 논문에서는 거리 기반 색인구조에서 k-최근접 질의를 효율적으로 처리하기 위해 데이터 분포 기반의 최적화 및 질의 처리 비용 기반 최적화 기법을 제안한다.