标准烟叶数据库的图像检索

Image Retrieval of Standard Tobacco Leaf Database

提出把图像的形状特征、颜色特征和纹理特征相结合的方法进行基于内容的图像检索。通过对人的视觉检索过程的研究 ,提出把数据库中的图像依次分别按形状特征、颜色特征和纹理特征分级聚类的方法 ,既符合人的视觉特点又大大提高了检索效率。

This paper deals with efficient retrieval of images from standard tobacco leaf database. Content based image retrieval is an important study field in multimedia database. In an image database, image retrieval is usually based on color histograms, but this method can not preserve the spatial information in the image. Unlike previous approaches that concentrates on extracting a single concise feature, our technique combines features that represent shape, color and texture in images. This method can either preserve statistical features or spatial information in the image. We brought forward a classification clustering method that clustered images according to shape, color and texture feature in turn. This method can greatly improve the retrieval efficiency. We adopted this method to retrieve images in ours standard tobacco leaf database, which showed that this approach achieved a better result.