Consumer photo management and browsing facilitated by near-duplicate detection with feature filtering

Wei-Ta Chu and Chia-Hung Lin

Department of Computer Science and Information Engineering,
National Chung Cheng University,
Taiwan

Abstract

Near-duplicate detection techniques are exploited to facilitate representative photo selection and region-of-interest (ROI) determination, which are important functionalities for efficient photo management and browsing. To make near-duplicate detection module resist to noisy features, three filtering approaches, i.e., point-based, region-based, and probabilistic latent semantic (pLSA), are developed to categorize feature points. For the photos taken in travels, we construct a support vector machine classifier to model matching patterns between photos and determine whether photos are near-duplicate pairs. Relationships between photos are then described as a graph, and the most central photo that best represents a photo cluster is selected according to centrality values. Because matched feature points are often located in the interior or at the contour of important objects, the region that compactly covers the matched feature points is determined as the ROI. We compare the proposed approaches with conventional ones and demonstrate their effectiveness.

Dataset

chlin_dataset.zip (32.4 MB)

Corresponding to Section 4.1 and Figure 9 [1], we provide training data for point-based filtering, region-based filtering, and pLSA-based filtering.
(1) Traing data for point-based filtering, region-based filtering, and pLSA-based filtering are in "point_based_traing", "region_based_traing", and "plsa_based_traing"
(2) Each directory consists of two sub-directories -- "arch" and "nature", which includes images of artifical objects and nature scene, respectively.
(3) The "sift" sub-directory stores SIFT features corresponding to each image, in the presentation of D. Lowe's definition. We extract SIFT features by Rob Hess's library: http://web.engr.oregonstate.edu/~hess/
(4) The directory "visword" stores visual words for artifical features and natural features. In each of these files, each row represents a visual word.

Citation

  1. W.-T. Chu and C.-H. Lin, "Consumer Photo Management and Browsing Facilitated by Near-Duplicate Detection with Feature Filtering," Journal of Visual Communication and Image Representation, vol. 21, no. 3, pp. 256-268, 2010.
  2. W.-T. Chu, C.-H. Lin, and J.-Y. Yu, "Feature Classification for Representative Photo Selection," Proceedings of ACM Multimedia Conference, pp. 509-512, 2009.
  3. W.-T. Chu and C.-H. Lin, "Automatic Selection of Representative Photo and Smart Thumbnailing Using Near-Duplicate Detection," Proceedings of ACM Multimedia Conference, pp. 829-832, 2008.

Any suggestion please send to