Personalized social image organization, visualization, and querying tool using low- and high-level features

Abstract

The purpose of this study is to create a software system to facilitate the organization of and searching for social images acquired from social sites on the Web (such as Facebook or Flikr), taking into account the images' features as well as user preferences. To achieve our goal, we design a solution based on image clustering, grouping together images sharing similar semantic and visual features, to simplify their organization and querying. This requires low-level and high-level image feature extraction and processing, where: low-level features represent color, texture, and shape image descriptors, whereas high-level features consist of textual descriptors extracted from image annotations and surrounding texts. Our system consists of modular components for: i) feature extraction and representation (low-level and high-level), ii) partitional image clustering (initial clustering phase executed when the user first connects to the system), iii) incremental clustering (updating clusters produces in the previous phase by processing newly published images), iv) fast image querying (using features of cluster representatives), and v) personalized images/search results visualization (using various user-chosen cluster display techniques). Preliminary experiments highlight the efficiency and practicality of our tool.