Content-based Image Retrieval Technology Research

Establishing images' indexes is one of the key problems in image retrieval. At present, the widely used and developed method is text-based image retrieval (TBIR), which is limited at objectivity sufficiency and automation in index establishing and representations. In response to these limitations, content-based image retrieval (CBIR) becomes a more intelligent way.In this paper, color and texture is used as image features. Image is retrieved based on global features and sub-image is retrieved based on local features.In color-based image retrieval, this paper proposes a two-step method. The six-box color histogram, which is obtained by color segmentation in HSV color space, is used for the coarse image retrieval. Then, the exact retrieval is accomplished by combining the cumulative color histogram. The experimental results show that the proposed method is better than single-step method.In texture-based image retrieval, this paper discusses Laws texture energy and statistical-based gray co-occurrence matrix as model to describe the feature. The features are represented compactly by computing the energy, entropy, contrast and uniformity of the co-occurrence matrix.In addition, this paper discusses the extraction and indexing methods in image region retrieval. Different from image segmentation, this paper actualizes regional extraction and indexing based on pre-defined color sets. The experiments show that the proposed method avoids the normally problems in image segmentation, and remedies the lost of color position information of color histogram based method.Finally, this paper designs and accomplishes an image retrieval system framework based on the Access database.