Bag of visual words approach for image retrieval using color information

Abstract

Content-based image retrieval has attracted researcher's attention due to the rapid growing of images database. Many approaches for content based image retrieval have been proposed but completely satisfactory result has not yet been obtained. In this paper a new approach for content-based image retrieval using the bag of visual words (BoW) has been proposed. In this approach, two separate visual vocabularies are produced based on the scale invariant feature transform (SIFT) and hue descriptor. Traditional SIFT BoW model do not utilize any color information in image description. We incorporate hue descriptor in our method which adds complementary information to BoW representation of images and increase the overall performance of retrieval. The quantization in BoW model is an essential step that has a great impact on the final result but it produces loss in quantization. To overcome this unfavorable effect, soft assignment is used in our proposed approach in which each feature vector can be assigned to multiple nearest visual words based on some weighting function. Experimental results on benchmark dataset showed that the proposed approach has a superior performance with respect to other methods.