摘要
全局颜色直方图是基于内容的图像检索中使用最广泛的特征之一。然而 ,它不包含颜色的空间分布关系 ,因此不同的图像可能具有同样的颜色直方图。采用图像分块的局部直方图方法考虑了颜色的位置信息 ,但却不具有旋转、平移不变性。文中利用分块颜色矩 ,借助带权二部图的最小完美匹配提出了一种新的相似度量方法 ,既考虑了颜色的空间分布关系 ,又具有旋转、平移不变性。
Global Color histogram is often used in content based image retrieval. But this method cant contain the spatial features. So different images may have the same color histogram. The local histogram partitioning an image into regions takes the spatial features into account, but fails to find rotated or translated images that are similar to query images. In this paper, a new method for image retrial using color moments of partitions and minimum cost perfect matching for a weighted bipartite graph based on image segmentation is proposed. This method of similarity measurement takes into account both the spatial features and the invariance of rotation and translation. Finally experiments show that the improvement of similarity measurement based on local color features is very effective.
出处
《计算机应用》
CSCD
北大核心
2004年第7期47-49,共3页
journal of Computer Applications
