基于梯度基元聚合矢量的图像检索算法*

针对图像检索中多特征融合问题,提出了一种基于梯度基元聚合矢量的图像检索算法。该算法在改进的HSV颜色空间计算边缘梯度,通过定义的基元模板扫描梯度图像,生成梯度基元图像,将基元和非基元像素分别组合成聚合和非聚合像素集合;最后利用颜色自相关图算法对上述两个集合提取特征矢量,实现了融合颜色、形状、纹理和空间信息等多特征的图像检索。实验结果表明,该算法能够融合颜色、形状、纹理和空间信息,有效地提高了基于内容的图像检索的查准率和查全率。

Image retrieval based on gradient texton coherence vector

This paper proposed a new image retrieval algorithm based on gradient texton coherence vector for fusion of multi-features. The algorithm computed the edge gradient first in the modified HSV color space, and then gained gradient texton map by scanning the gradient image through the special texton types. The texton pixels were combined into the coherence set, the other pixels were the non-coherence set. At last,it represented the feature vector of the image retrieval by color auto-correlogram in two sets, and realized the image retrieval based multi-features which fused color, shape, texture and spatial information. Experimental results show that the proposed algorithm can combine color, texture, shape and spatial characteristic effectively, and have valid precision and recall.