对深度学习领域的稠密卷积网络（dense convolutional network,DenseNet）进行改进，提出了一种嵌套网络模型下的相似图像检索方法。该方法主要通过嵌入压缩和激励网络（squeeze-and-excitation network,SENet），调整原DenseNet网络结构，优化特征提取模块，从而提高图像检索的准确率。在整个深度学习的过程中，给图像特征通道设置合理的权值，抑制图像中的无效特征，能够进一步提高图像的检索速度。实验结果表明，所提算法能够加强图像有效特征的传递，无论从精度和速度方面均可得到较好的图像检索结果。
A similar image retrieval method based on nested network model is proposed by improving the traditional dense network (DenseNet) which is a common deep-learning method. First, the proposed algorithm optimizes feature retrieving blocks by embedding squeeze-and-excitation network (SENet) into the original DenseNet and adjusting the structure of the DenseNet, so as to improve the accuracy of image retrieval. Second, the algorithm achieves the speed-up of image retrieval by setting proper weight for each feature channel of the image, thus suppressing the processing time of those invalid features. Experimental results show that the algorithm can strengthen the transmission of effective image features and improve the accuracy of image researching results effectively.
[1] 李志文,王洪添,于治楼.图像检索中基于复杂图像特征的相似度计算[J].信息技术与信息化, 2008(4):40-43. Li Z W, Wang H T, Yu Z L. Similarity calculation based on complex image features in image retrieval[J]. Information Technology and Informatization, 2008(4):40-43.(in Chinese)
[2] 张丽.基于颜色和纹理特征的图像检索技术研究[D].南京:南京邮电大学, 2017.
[3] Luis L, Cesar P. Performance evaluation for the hash generation phase of a democratic blockchain[J]. International Journal of Internet Technology and Secured Transactions, 2020, 10(3):286-303.
[4] Shu X, Wu X J. A novel contour descriptor for 2D shape matching and its application to image retrieval[J]. Image and Vision Computing, 2011, 29(4):286-294.
[5] Datta R, Joshi D, Li J, Wang J Z. Image retrieval:ideas, influences, and trends of the new age[J]. ACM Transactions on Computing Surveys, 2008, 40(2):1-66.
[6] Rubner Y, Tomasi C, Leonidas J. The earth Mover's distance as a metric for image retrieval[J]. International Journal of Computer Vision, 2000, 14(2):99-121.
[7] Akakin H C, Gurcan M N. Content-based microscopic image retrieval system for multi-image queries[J]. IEEE Transactions on Information Technology in Biomedicine, 2012, 16(4):758-769.
[8] Tolias G, Sicre R, Jegou H. Particular object retrieval with integral max-pooling of CNN activations[C]//2016 IEEE Conference on Computer Vision and Pattern Recognition, 2016:1-12.
[9] Simonyan K, Zisserman A. Very deep convolutional networks for large-scale image recognition[J]//Computer Science, 2014:1-14.
[10] He K M. Deep residual learning for image recognition[C]//2015 IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2015), 2015:770-778.
[11] Zhuang L, Gao H, Kilian Q. Densely connected convolutional networks[C]//IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2017), 2017:4700-4708.
[12] Jie H, Li S, Gang S. Albanie S. Squeeze-and-excitation networks[C]//IEEE Conference on Computer Vision and Pattern Recognition (CVPR 2017), 2017:7132-7141.
