| Summary: | 碩士 === 國立高雄第一科技大學 === 電腦與通訊工程所 === 90 === As the advances of the Internet, the demand of storing multimedia information (such as text, image, audio, and video) has increased. And the multimedia retrieval and search is more and more important. Traditionally, textual features such as filenames, captions, and keywords have been used to annotate and retrieve images. As it is applied to a large database, the use of keywords becomes not only cumbersome but also inadequate to represent the image content. Therefore, many content-based image retrieval system have been proposed to solve this problem.
In this thesis, we propose an efficient object-based image retrieval method by using a fast K-NNR search algorithm which is designed according to the triangle inequality principle. The computational complexity of the traditional histogram-based image retrieval which is also improved by fast K-NNR search method is high due to the usage of the high-dimensional histogram and the lack of the indexing structure. Furthermore, a new indexing structure for the proposed object-based image retrieval technique is also proposed in this study. A special attention to object segmentation which combines the moment-preserving edge detection and the region-growing techniques is paid in this thesis. Finally, an object-based similarity metric is also proposed for query processing.
Experimental results show that the proposed image retrieval method is effective and superior to other methods in terms of overall computational complexity. Applying to a very large image database, the performance of the proposed method can be sustained.
|
|---|
