Large Image Collections – Comprehension and Familiarization by Interactive Visual Analysis

Large size and complex multi-dimensional characteristics of image collections demand a multifaceted approach to exploration and analysis providing better comprehension and appreciation. We explore large and complex data-sets composed of images and parameters describing the images. We describe a novel approach providing new and exciting opportunities for the exploration and understanding of such data-sets. We utilize coordinated, multiple views for interactive visual analysis of all parameters. Besides iterative refinement and drill-down in the image parameters space, exploring such data-sets requires a different approach since visual content cannot be completely parameterized. We simultaneously brush the visual content and the image parameter values. The user provides a visual hint (using an image) for brushing in addition to providing a complete image parameters specification. We illustrate our approach on a data-set of more than 26,000 images from

Flickr

. The developed approach can be used in many application areas, including sociology, marketing, or everyday use.