Conference Paper

Adding Affine Invariant Geometric Constraint for Partial-Duplicate Image Retrieval.

DOI: 10.1109/ICPR.2010.212 Conference: 20th International Conference on Pattern Recognition, ICPR 2010, Istanbul, Turkey, 23-26 August 2010
Source: DBLP

ABSTRACT The spring up of large numbers of partial-duplicate images on the internet brings a new challenge to the image retrieval systems. Rather than taking the image as a whole, researchers bundle the local visual words by MSER detector into groups and add simple relative ordering geometric constraint to the bundles. Experiments show that bundled features become much more discriminative than single feature. However, the weak geometric constraint is only applicable when there is no significant rotation between duplicate images and it couldn't handle the circumstances of image flip or large rotation transformation. In this paper, we improve the bundled features with an affine invariant geometric constraint. It employs area ratio invariance property of affine transformation to build the affine invariant matrix for bundled visual words. Such affine invariant geometric constraint can cope well with flip, rotation or other transformations. Experimental results on the internet partial-duplicate image database verify the promotion it brings to the original bundled features approach. Since currently there is no available public corpus for partial-duplicate image retrieval, we also publish our dataset for future studies.

  • [Show abstract] [Hide abstract]
    ABSTRACT: Recently, feature grouping has been proposed as a method for improving retrieval results for logos and web images. This relies on the idea that a group of features matching over a local region in an image is more discriminative than a single feature match. In this paper, we evolve this concept further and apply it to the more challenging task of landmark recognition. We propose a novel combination of dense sampling of SIFT features with interest regions which represent the more salient parts of the image in greater detail. In place of conventional dense sampling used in category recognition that computes features on a regular grid at a number of fixed scales, we allow the sampling density and scale to vary based on the scale of the interest region. We develop new techniques for exploring stronger geometric constraints inside the feature groups and computing the match score. The spatial information is stored efficiently in an inverted index structure. The proposed approach considers part-based matching of interest regions instead of matching entire images using a histogram under bag-of-words. This helps reducing the influence of background clutter and works better under occlusion. Experiments reveal that directing more attention to the salient regions of the image and applying proposed geometric constraints helps in vastly improving recognition rates for reasonable vocabulary sizes.
    The Visual Computer 06/2013; 29(6-8). · 0.91 Impact Factor
  • Source
    [Show abstract] [Hide abstract]
    ABSTRACT: Discovering common visual patterns (CVPs) between two images is a difficult and time-consuming task, due to the photometric and geometric transformations. The state-of-the-art methods for CVPs discovery are either computationally expensive or have complicated constraints. In this paper, we formulate CVPs discovery as a graph matching problem, depending on pairwise geometric compatibility between feature correspondences. To efficiently find all CVPs, we propose a novel framework which consists of three components: Preliminary Initialization Optimization (PIO), Guided Expansion (GE) and Post Agglomerative Combination (PAC). PIO gets the initial CVPs and reduces the search space of CVPs discovery, based on the internal homogeneity of CVPs. Then, GE anchors on the initializations and gradually explores them, to find more and more correct correspondences. Finally, to reduce false and miss detection, PAC refines the discovery result in an agglomerative way. Experiments and applications conducted on benchmark datasets demonstrate the effectiveness and efficiency of our method.
    Journal of Visual Communication and Image Representation 07/2013; 24(5):635–646. · 1.20 Impact Factor
  • [Show abstract] [Hide abstract]
    ABSTRACT: Many image retrieval and search systems take visual words as the key to represent images. Extracted feature descriptors are assigned to best matching visual words in the vocabulary. However, the visual word based retrieval method is not satisfactory in real applications as the descriptive ability of visual vocabulary is not strong enough to describe image local features. In this paper, we investigate on the problem of the descriptive ability of visual vocabulary. There are mainly two contributions in this paper. Firstly, we make a comprehensive analysis on the retrieval performance of standard bag-of-features (BOF) method by using different codebooks. The codebook is generated on one dataset and used for retrieval on the same dataset and on different datasets, which is called ''homology codebook'' and ''non-homology codebook'', respectively. Experimental results show that there exists performance drop when a non-homology codebook is used for retrieval on one dataset, compared to the homology codebook. This phenomena is usually neglected by most of the retrieval tasks and systems. Secondly, in order to abate the influence of non-homology codebook, a weighting based method is proposed to balance the descriptive ability of codebook for different datasets. Experimental evaluations show that the proposed method outperforms standard BOF method, and it can prevent the performance drop to some extent when non-homology codebooks are used.
    Neurocomputing. 11/2013; 119:478-488.

Full-text (2 Sources)

Available from
Jun 2, 2014