融合颜色词袋特征的视觉词汇树图像检索

针对由图像灰度空间产生的传统词袋模型SIFT特征无法体现图像的颜色信息的问题,提出了一种融合颜色特征的视觉词汇树来对图像进行描述。提取SIFT特征并建立词汇树,获取图像的SIFT表示向量。利用K-means方法对图像库中的所有图像的HSV值进行聚类,获得基于HSV空间的颜色词袋表示向量,避免了传统颜色直方图方法所带来的量化误差。将SIFT特征与颜色词袋特征进行融合,完成了图像的全局特征和局部特征的融合。然后,计算融合特征的相似度,将相似度从高到低排序,完成图像检索。为了验证本方法的有效性,选择Corel图像库对算法性能进行实验分析,从主观评价和客观评价标准分别进行评价,并与传统方法进行了对比。结果表明,特征融合的检索性能与单一特征方法相比有较大提高。特征融合方法的平均检索查准率和查全率-查准率等评价指标,对比传统方法均有不同程度提高。

Image retrieval based on visual vocabularytree fusing color wordbag feature

In the traditional bag of word model, the SIFT features are extracted from the gray space of image, which cannot reflect the color information of image. To solve this problem, we propose to use a visual vocabulary tree vector that fuses the color feature to represent image contents. SIFT features are extracted and the vocabulary tree is built to obtain the SIFT features of images.The K means method is used to cluster the HSV values of all images in the image library so as to obtain the representation vector of the color word bag based on the HSV space, there by avoiding the quantization error brought by the traditional color histogram method.The fusion of SIFT features and color word bag features completes the fusion of global and local features of the image.Finally, by calculating the similarities of the fusion features and sorting them from high to low,the image retrieval is completed. In order to validate the effectiveness of the proposed method, we choose Corel image database to analyze the performance of the algorithm, evaluate it from subjective evaluation and objective evaluation criteria, and compare it with the traditional method. The results show that,compared with the single feature method, the proposal improves the retrieval performance of feature fusion. The average retrieval precision and the recall ratio of the feature fusion method are all improved to some extent.