改进视觉词袋模型的快速图像检索方法

Fast Image Retrieval Method Using Improved Bag of Visual Words Model

视觉词袋模型在基于内容的图像检索中已经得到了广泛应用,传统的视觉词袋模型一般采用SIFT描述子进行特征提取.针对SIFT描述子的高复杂度、特征提取时间较长的缺点,本文提出采用更加快速的二进制特征描述子ORB来对图像进行特征提取,建立视觉词典,用向量间的距离来比较图像的相似性,从而实现图像的快速检索.实验结果表明,本文提出的方法在保持较高鲁棒性的同时,明显高了图像检索的效率.

Bag of visual words model based on content-based image retrieval has been widely used, traditional bag of visual words model generally uses the SIFT descriptors for feature extraction. In view of the high complexity of SIFT descriptors and the long time of feature extraction, this paper proposes to use a faster binary feature descriptor ORB for the image feature extraction, creating visual dictionary, using the distance between two vectors to compare the image similarity, so as to achieve fast image retrieval. Experimental results show that the method proposed in this paper can improve the efficiency of image retrieval obviously, while maintains a relatively high robustness.