Academic
Publications
Re-ranking of web image search results using a graph algorithm

Re-ranking of web image search results using a graph algorithm,10.1109/ICPR.2008.4761472,Hilal Zitouni,Sare Gul Sevil,Derya Ozkan,Pinar Duygulu

Re-ranking of web image search results using a graph algorithm   (Citations: 8)
BibTex | RIS | RefWorks Download
We propose a method to improve the results of image search engines on the Internet to satisfy the users who desire to see the relevant images in the first few pages. The results of the text based systems, that use only the accompanied text of the images, are re-ranked by incorporating the visual similarity of the resulting images. We observe that, in general, together with many unrelated ones, the result of text based systems include a subset of correct images, and this set is the largest most similar one compared to other possible subsets. Based on this observation, we present the similarities of all the images in a graph structure, and find the largest densest component of the graph, corresponding to the largest set of most similar subset of images. Then, to re-rank the results, we give higher priority to the images in the densest component, and rank the others based on their similarities to the images in the densest component. The experiments carried out on 10 category of images from (4) promise the success of our method over Google ranking. In this study, we propose a method to satisfy the users of image search engines by re-ranking the results of text based systems using visual information. In our approach, with the assumption that there will be a large set of visually similar images relevant to the query in the set of all resulting images, we find the largest set of most similar images and place them in the earlier pages.
Cumulative Annual
View Publication
The following links allow you to view full publications. These links are maintained by other sources not affiliated with Microsoft Academic Search.
    • ...Clustering based approaches [3, 34] typically assume that the set of relevant images form the largest cluster amongst the top ranked results...
    • ...Note that the number of queries in our development set alone is comparable to the previous image re-ranking methods – between 10 and 20 in [3, 4, 26, 30, 34], 26 in [31], and 45 in [8]...

    Vidit Jainet al. Learning to re-rank: query-dependent image re-ranking using click data

    • ...In 36, the authors presented similarities of all images in a graph structure and found the densest component that corresponds to the largest set of most similar subset of images...

    Chaoran Cuiet al. A hybrid relevant-diverse approach for image re-ranking with multiple ...

    • ...Besides image classification based approach, there are some other approaches to improve text based web image search by exploiting image content, such as user interactive methods [8-9], image similarity based methods [10-15], multiple search engines based method [16], and etc...
    • ...Zitouri et al. [12] presented similarities of all images in a graph structure and find the densest component that corresponds to the largest set of most similar subset of images...

    Yuchai Wanet al. Online image classifier learning for Google image search improvement

    • ...In literature, there are many works about visual re-ranking based on different schemes, such as clustering-based [15][21], classific ation-based [19] and the graph-based [8][16][17][18][20][22][24]...
    • ...Currently, the similarity is mainly estimated based on image low-level features: global feat ures [17][18][24], such as color moments and Gabor feature, and local features [8][20][22][24], such as SIFT (Scale Invariant Feature Transform) [2]...

    Wengang Zhouet al. Latent visual context analysis for image re-ranking

    • ...Recently, many reranking methods have been proposed, including the classification-based [14], clusteringbased [1, 4] and graph-based [5, 16, 12, 7] methods...
    • ...Currently, the similarity is predominantly estimated based on images’ low-level features: either global features [16, 12], such as color moments [9] and texture [2], or local features [5, 7], such as SIFT (Scale Invariant Feature Transform) features [8]...
    • ...For example, in [16], the authors found that the local feature gives good results for queries like “zebra”, “car” and “guitar” while the global feature shows effectiveness for queries like “horse” and “bikes”...

    Li Wanget al. Query aware visual similarity propagation for image search reranking

Sort by: