基于形状特征分划分的树状区域匹配图像检索

提出了一种依据形状特征划分的树状区域匹配的图像检索算法，其思想是首先基于形状特征将图像发为若干给定浓度的树状关系区域，再结合所划分区域的结构关系及其颜色特征进行匹配检索。通过2个检索粗过滤条件，显著提高了检索响应速度，应用此算法，图像中对象的几何变换（移动、旋转、比例缩放）不会影响到查询结果，实验测试的结果表明，与一般的颜色直方图匹配算法相比，该算法大大提高了检索效率。

Regions with Tree Relationships Segmented on Shape Features Matching for Image Retrieval

An algorithm for color image retrieval by regions with hierarchical tree relationships matching is presented. First segment an image into several regions with relationship tree structure in the degree presented beforehand based on shape features. Then by combining the information of the relationship tree structure and color histogram of the regions images can be retrieved more accurately. The response time of retrieval is strongly reduced by two rough filter conditions set in the algorithm. Fundamental geometry transformation, such as move, rotation, scale, of the objects in images does not affect retrieval results by using the algorithm. The efficiency of retrieval by using this algorithm is highly increased compared with that of retrieval by using normal color histogram matching as shown by the retrieval experiments with real image database.