Int J Performability Eng ›› 2018, Vol. 14 ›› Issue (2): 223-231.doi: 10.23940/ijpe.18.02.p4.223231

• Original articles • Previous Articles     Next Articles

A Novel Image Retrieval Method with Saliency Feature Vector

Junfeng Wua, b, Wenyu Quc, *, Zhiyang Lid, and Changqing Jie   

  1. aSchool of Computer Science and Technology, Tianjin University, Tianjin, 300072, ChinabSchool of Information Engineering, Dalian Ocean University, Dalian, 116023, China,cSchool of Software, Tianjin University, Tianjin, 300072, ChinadSchool of Information Science and Technology, Dalian Maritime University, Dalian, 116026, China eSchool of Physical Science and Technology, Dalian University, Dalian, 116023, China

Abstract:

In the past few years, image retrieval has been one of the research focuses in the field of computer vision. For most retrieval methods, the accuracy of the retrieval results mainly depends on the extracted feature vectors. But, the foreground and the background in the images are not distinguished for most methods. It is obvious that these methods are not in accordance with the visual characteristics of the human eye. In this paper, salient objects are extracted from images in order to improve the pertinence of feature vector extraction. The paper utilizes a spatial pyramid model to divide the image into different parts with different scales. The feature vectors extracted in different scale are connected. Then, the saliency map and saliency score are used to rebuild the joint vector. Each feature vector is assigned different weighted values according to its different location in the image and scale. Finally, the newly constructed feature vectors are used to measure the similarity between images. In order to test the effectiveness of the algorithm, we evaluate our method on the SIMPLIcity dataset and Stanford dataset. Experimental results show that the proposed method has a great improvement in both accuracy and efficiency.