Abstract: The paper proposed a new method for document images retrieval based on texture spectrum descriptor. This algorithm firstly segmented the text area by document image characteristics, and then described the texture spectrum based on the text character edge, and computed the texture spectrum histogram. Compared with direct use of gray histogram image retrieval, this algorithm possesses high distinguish degree. The experimental results show that this method has high retrieval precision and resists cutting or rotating operations, and is suitable for document image retrieval.

Key words: document image, 　image segmentation, 　image texture, 　CBIR, 　image matching

摘要： 新算法首先根据文档图像的特点分割图像文本区域,并将文档图像中字符的边缘信息使用纹理谱进行描述,计算纹理谱图像的直方图。相对于直接使用灰度直方图进行图像检索,该算法具有更好的区分度。实验结果表明,该方法具有很高的查准率,并对剪切、旋转操作表现出很好的稳定性,适合文档图像检索。

关键词: 文档图像, 图像分割, 图像纹理, 基于内容的图像检索, 图像匹配