On Image Retrieval Using Salient Regions with Vector-Spaces and Latent Semantics

Abstract

Abstract. The vector-space retrieval model and Latent Semantic Indexing approaches to retrieval have been used heavily in the field of text information retrieval over the past years. The use of these approaches in image retrieval, however, has been somewhat limited. In this paper, we present methods for using these techniques in combination with an invariant image representation based on local descriptors of salient regions. The paper also presents an evaluation in which the two techniques are used to find images with similar semantic labels.

A Correlation Approach for Automatic Image Annotation

“…[8] has kindly provided us with the post processed data. SIFT descriptors were computed from the images and then clustered using the batch k-means clustering algorithm with random starting points in order to build a vocabulary of 'visual' words [7]. Each image in the entire data-set then had its feature vectors quantised by assigning the feature vector to the closest cluster.…”

“…We apply the TFIDF on the image SIFT descriptors as they were post processed as to mimic the concept of words (SIFT descriptors) in documents (images), the pseudo-details of this procedure are given in the following section and further information can be found in [7]. In the experiments results section we compare the application of TFIDF on the visual terms as well as keeping them as frequency vectors.…”

“…Recently in [7,8], it was suggested that methods that use region-based image descriptors generated by automatic segmentation or through fixed shapes may lead to poor performance, as regularly used rectangular regions image descriptors are not robust to a variety of transformations such as rotation. They have suggested using Scale Invariant Feature Transformation (SIFT) [9] feature, which are scale invariant, and utilising them as 'visual' terms in a document.…”

<title>MapSnapper: engineering an efficient algorithm for matching images of maps from mobile phones</title>

“…In particular, the use of salient regions for robust image matching and retrieval [5][6][7][8][9][10][11][12] has been exploited. Our own previous research on mobile retrieval within a museum setting 3 was used as a basis for the design of the system architecture.…”

A Linear-Algebraic Technique with an Application in Semantic Image Retrieval

“…Each of the images was indexed using 'visual' terms from quantised local SIFT descriptors about interest points picked from peaks in a difference-of-Gaussian pyramid [9,10]. The size of the visual vocabulary was fixed to 3000 terms [10].…”