Abstract: In order to overcome the limits of image segmentation, a new method is presented in image retrieval process using salient points matching technique. A region of an image is selected manually as the query image. Then, the Harris color salient points are extracted and expressed using corresponding color characteristics. Finally, the salient points matching technique is used to realize the region-based image retrieval process. Experiments show that this method is robust to image illumination changes and geometrical transformations, and it can also improve the retrieval performance effectively.

Key words: Salient points; Region-based image retrieval; Image index; Similarity measure

摘要： 为了克服图像分割技术的限制，提出了一种基于特征点匹配技术的图像检索算法。手工提取图像中的一块区域作为查询图像，然后使用Harris 算子提取彩色特征点，并用相应的颜色特征对特征点进行表示，最后利用特征点匹配技术实现区域图像的检索。实验表明，该方法对于图像的亮度变化和几何变换具有很强的鲁棒性，可以有效提高检索准确率。

关键词: 特征点；区域图像检索；图像索引；相似性度量