How small should a document thumbnail be?

Viewing document images on small devices is a challenge. When showing a region of a page for reading on a small display a page overview is generally lost. If a page overview is desired, typically a low resolution version of the image fitting a small array of pixels a thumbnail is provided. Whereas the readability of text in thumbnails is often lost, document layout information may or may not be preserved. Preserving document layout information in thumbnails is the goal of this paper. We derive models for controlling the preservation of document layout information in thumbnails by determining the size of a thumbnail depending on the layout content of the document. The downsampling factor for a document image will depend on its layout information, such that layout units will be visually separate after scaling. The link between scaling factors and document layout information is created through novel models, White Space Graphs and White Space Trees. These models enable control over enhancement and suppression of document layout structures during scaling. Minimal scaling factors can be derived that assure visual separability of a controlled set of layout units after scaling. Those scaling factors depends on the document content as well as user and display specifications.