植基於距離測量之影像檢索系統

A Novel Content-Based Image Retrieval System Based on Distance Measure Approach

林嘉文(Chia-Wen Lin)

指導教授 ： 黃慧鳳 ； 黃國峰

國立臺中科技大學/資訊與流通學院/資訊工程系碩士班/碩士(2013年)

https://doi.org/10.6826/NUTC.2013.00084

摘要

以內容為基底的影像檢索(content-based image retieval)是近二、三十年來相當受到注意的影像處理技術之一。許多研究者透過萃取影像的顏色、紋理、形狀或空間位置得到影像的特徵，再利用特徵之間的比對，來找到資料庫裡與輸入影像最類似的數張影像。研究者們鑽研出許多不同的特徵萃取方法，以及各種不一樣的距離比對方式，希望能夠簡單又快速的獲取更佳的影像檢索準確率。
在這篇論文中，我們提出一個新的距離比對方式，利用第一次檢索過後的結果，找出參照特徵，再次進行檢索並對結果排序(re-rank,簡稱RRK)，能夠非常輕易的提升影像檢索的效率。通過我們的實驗，結果證明了我們的方法是相當有效的，而且我們相信這個方法可以被應用在其他的影像檢索技術中，提升其影像檢索效能。

關鍵字

影像檢索系統 ； CBIR ； HSV色彩空間 ； 灰階共現矩陣 ； 距離比對

並列摘要

For the last two dacades, content-based image retrieval (CBIR) is a very popular topic in image processing area. Researchers aim to extract features of color, texture, shape or position information from an image and then via a distance measure method, they retrieve images most similar to the query image from a particular image database. A lot of content-based image retrieval methods have been proposed and the performance of each mothod has been demonstrated in each proposed paper.
In this paper, we proposed a novel distance meature approach (re-rank,shorted as RRK) for CBIR system. First, we get the retrieved images from our simple CBIR system. Then we used these images to set reference feature. Finally, we retrieved again using the referenced feature and obtained more precise result. According to the experimental results, the proposed method is simple and efficient. Furthermore, the proposed method can be applied in other CBIR system easily.

並列關鍵字

CBIR ； content-based image retrieval ； HSV ； gray level co-occurrence matrix ； image matching

參考文獻

1. Liu, G.-H., Li, Z.-Y., Zhang,L. and Xu,Y., Image retrieval based on micro-structure descriptor. Pattern Recognition, 2011. 44(9): p. 2123-2133.

2. Yue, J., Li, Z., Liu, L. and Fu, Z., Content-based image retrieval using color and texture fused features. Mathematical and Computer Modelling, 2011. 54(3–4): p. 1121-1127.

3. Wang, X.-Y., Yu, Y.-J. and Yang, H.-Y., An effective image retrieval scheme using color, texture and shape features. Computer Standards & Interfaces, 2011. 33(1): p. 59-68.

4. Singha, M. and Hemachandran, K., Content Based Image Retrieval using Color and Texture. Signal & Image Processing, 2012. 3(1).

5. Smith, J.R. and Chang, S., Transform Features for Texture Classification and Discrimination in Large Image Databases. Proceeding, in IEEE International Conference on Image Processing. 1994. p. 407-411.