Survey on Visual Word Generation and Their Matching Techniques

Abstract

Content-Based Image Retrieval (CBIR), a technique which uses visual contents to search images from large scale image databases according to user’s interests. Contents of image can be global features such as color, shape and texture or local features such as SIFT (Scale Invariant Feature Transform) features. Both types are complement for each other. Bag of visual words (BoVW) is a widely used approach in most of content based image retrieval. This approach is based on local features of images which forms clusters based on similarity of these features. These formed clusters are used for matching two similar images. This survey paper mainly concentrate on clustering of local features i.e. visual word generation and matching of similar visual words based on their characteristics. This paper surveys on recent studies on visual word generation and matching.
Key Words: Bag of Visual Words (BOW), CBIR, KD-tree