用颜色特征进行基于形状的检索

现有的多媒体数据库系统中对图片的检索基本上是基于描述关键字的，近几年开始研究基于内容的检索。基于内容的图像检索技术大多采用颜色特征。对于以形状为主的图片来说，基于颜色的检索就没有任何意义。但对真彩色图像提取形状特征是比较困难，且大部分用形状特征来匹配图形的算法，对图形的旋转、伸缩等不具有鲁棒性。该文采用颜色特征来获取图像中物体的形状特征，并基于此进行图像形状匹配算法。同时通过对算法的改进，使图像的大小变化和旋转以及轻微的光照变化不影响检索结果。实验结果表明这种方法改善了检索效果。

Shape-Based Retrieval Using Color Feature

Recently in most Multimedia database, the methods used in image retrieval are mainly based on described keywords, now content--based retrieval are coming into research and color feature is mostly used. But tOthose pictures which are represented by object's shape, color--based retrieval is meaningless. As it is difficuh toget the shape feature from a color image, and also most image--match algorithms based on shape are not robust torotation and scale of an image. In this paper, the objecta's edge feature are obtained from image's color, and usingthese color feature to fulfill the shape--based algorithm. Also by improving the algorithm, this method is effective toimage's rotation and scale, even the slightly change of illumination. The experiment results indicate that thismethod improve the retrieval effect.