以图像颜色聚合向量为基础，并结合图像显著特征，提出了一种基于加权颜色聚合向量的图像检索方法。首先，提取图像的显著性图，并进行归一化处理，得到加权矩阵；然后，对图像进行颜色聚合向量提取，并根据加权矩阵进行加权处理；最后通过计算两幅图像之间的加权颜色聚合向量相似度，进行图像检索。该方法既系统兼顾了图像的颜色分布特征和高层视觉特征，又具有较高的计算速度；实验结果证明，该算法的检索精度明显高于传统的基于颜色统计特征的检索精度。 By taking visual saliency into consideration ,a new image retrieval method based on color coherence vector is proposed in the paper .According to the saliency map of the image ,every pixel is assigned a weighting value . Thus ,the color coherence vector is calculated on those weighted pixels .This new vector reflects both the low-level color region distribution feature of the image ,and its high-level vision characteristics .The experiments verify this method above is more efficiently than traditional ways based on color histogram .
English Abstract:
By taking visual saliency into consideration, a new image retrieval method based on color coherence vector is proposed in the paper. According to the saliency map of the image, every pixel is assigned a weighting value. Thus,the color coherence vector is calculated on those weighted pixels. This new vector reflects both the low-level color region distribution feature of the image, and its high-level vision characteristics. The experiments verify this method above is more efficiently than traditional ways based on color histogram.
