Written in Java, Tapestry cycles through a database of hundreds of globally and ideologically dispersed online news URLs to locate and extract relevant news images. Individual images are analyzed and temporarily cataloged according to content (e.g., color content and histogram data). Like images cluster together, and big news stories with widespread coverage tend to create large clusters. Less "popular" stories create smaller clusters, and single stories are represented by singleton, outlier images.

Clustering and scaling accentuate the semantic value of image elements. Clusters of identical images immediately reveal image recycling by news sources, regionally and throughout the world. Clusters form, expand, and recede based on the ebb and flow of news stories.

In the example at right, the green squares represent individual images related to a story that is prevalent in much of the news media around the world; the display immediately conveys that the story is receiving coverage in North and South America, Europe, Asia, and Australia. The largest clusters indicate the most prominent stories, overall and within regions. In this case, the most prominent overall story is associated with images represented by green squares. The yellow squares represent a story with media focus primarily in eastern Europe, with some mention in North America and Africa. The red square represents an outlier story—one with coverage by only one news source, in only one region.

The Tapestry design employs three main functional components for loading URLs, parsing page content, and placing the extracted images on the display. External, editable text files contain parameters used to adjust the search scope and control image placement.