Efficient mobile landmark recognition based on saliency-aware scalable vocabulary tree

Abstract

In recent years, the Scalable Vocabulary Tree (SVT) has been shown to be effective in image recognition. However, in mobile landmark image recognition where the foreground is the landmark to be recognized while the background is cluttered, the current SVT framework ignores different local importance of image, hence restricting its performance. In this paper, we propose a new landmark recognition framework that can incorporate saliency information to improve the recognition performance relative to the baseline SVT method. Specifically, the saliency information is incorporated in three phases: image descriptor calculation, vocabulary tree generation, and image representation. We constructed a city-scale landmark dataset in Singapore, and the experimental results show that the proposed mobile landmark recognition by incorporating saliency information outperforms the baseline SVT recognition by about 9%.

Re-ranking approach to mobile landmark recognition

“…We propose a re-ranking method to improve the performance of mobile landmark recognition systems that incorporate the saliency information and the SVT image recognition [2].…”

“…Graph-Based Visual Saliency (GBVS) [9,2] is one representative process in visual saliency modeling. It consists of two steps: first, forming an activation map A on the certain feature map F, and then normalizing it in a way which highlights conspicuity locations of the saliency.…”

“…The server analyzes the information and sends the relevant information about the landmark image, such as its name and associated activities, to the user. There are many research papers on mobile landmark recognition, e.g., David et al [1], Kim et al [2], Rongrong et al [3], Tao Chen et al [4] Zhang et al [5], and Irschara et al [6]. In [7,2], where the scale invariant feature transform (SIFT) is used to transform the image data into scale-invariant coordinates relative to local features.…”

“…The recognition system is based on the saliency weighting of SIFT and SVT [2]. The reranking is applied to this system in the final stage to improve recognition performance.…”

Image retrieval by information fusion based on scalable vocabulary tree and robust Hausdorff distance

“…SVT approaches can be regarded as an extension of Bagof-Words (BoW) approaches since the visual words can be easily extended to tens of thousand at a logarithmic scale [10][11][12][13]. In a typical image retrieval framework [1], a scalable vocabulary tree (SVT) is generated by hierarchically clustering local descriptors.…”