Manipulation of visual information refers to the activities performed to restructure information presentation. Systems can be classified according to the techniques that control dynamics of the presentation and according to the granularity of the controlled components in the presentation.
Systems typically employ distortion, zoom, multiple degrees of freedom (DOF), tight-coupling and metaphor-based techniques to control dynamics (Figure ).
Distortion based techniques use a global transformation function that maps the original layout of the presentation to a another layout. Users' interaction typically results in changing the parameters of the transformation function, such as translating and scaling the function. Such interactions allow the users to examine different regions of the layout and also under different parameters. Graphical Fisheye Views , Hyperbolic Tree , Bifocal Displays  and Keahey et al.  demonstrate interesting distortion based techniques.
Zooming techniques allow the user to translate the view of the presentation for each dimension of the presentation by pan left/right (x-translation) and up/down (y-translation), and zoom in/out (z-translation) operations. Typically, current view only shows a portion of the whole presentation. Pad++  is an impressive zooming user interface for the exploration of very large information structures. In Pad++ objects are placed on a two-dimensional infinitely large layout at multiple scale factors. Users initially start with an overview of the information space where they can then zoom in on a region smoothly to see more details. It also provides semantic zooming, where objects are displayed at multiple levels of abstractions depending on the zoom factor. Other zoom-based interfaces are News in the Future , LifeLines  and Dynamic Queries .
Some systems use multiple degrees of freedom (i.e. translations and rotations on each dimension) to navigate or manipulate the view of the presentation. In the Harmony browser  users explore the three-dimensional information landscape by manipulating objects (translate, rotate, zoom) and by navigating (walk, fly, fly to, heads up). Similarly, users interact with the Information Cube  by thee-dimensional translations and rotations. Multiple degrees of freedom interaction is also typically found in Virtual Reality Modeling Language (VRML) Browsers. These systems typically require the use of special input devices for interaction than can handle multiple degrees of freedom - typically 6, or use mode-based interaction where lesser degrees of freedom devices can be used instead.
In tightly-coupled techniques user's action in one part of the presentation affects other parts dynamically. The Visible Human Explorer  interface demonstrate a pair of tight-couplings between two different views (e.g. coronal and axial) of a human body. Users explore the human body by sweeping the views through the body with smooth, rapid, visual feedback. LinkKit  is a prototype coordination toolkit that facilitates coordination for browsing a group of documents in multiple windows by allowing users to create on-demand associations. For example, users can create a tight-coupling between two documents for synchronous scrolling. The PDQ (Pruning with Dynamic Queries)  Tree-browser provides the users with an overview of the whole tree structure and a tightly-coupled separate detail view. When the user moves the field-of view box in the overview window, the corresponding view is updated in the detail view.
Metaphor-based techniques demonstrate real-life like dynamics on the presentation. For example, the WebBook and the Web Forager  demonstrate a number of such dynamics in a three-dimensional book-desk-shelf metaphor. Users can take a book from the shelves, open and ruffle its pages. All such dynamics are done through realistic three-dimensional animations. Mander et al.  examined the pile metaphor to support casual information organization where users can browse and reorganize piles of documents. Dynamics in the current windowing systems such as moving windows are also considered metaphor based because they resemble the movements on a desktop.
All of the above manipulation techniques can be applied at different granularities on the components of the presentation such as the pixel, object and structure level. Figure 1 demonstrates a magnification dynamics in each class of dynamics based on distance to the center of the image.
Space-based techniques consider entities in terms of the occupied space. Entities are composed of pixels, which are the unit of space-based operations. Thus, each pixel of an entity is manipulated independently with no regard to the entity it belongs to. Bifocal Displays  employs a space-based distortion technique to magnify a rectangular focus region on a two-dimensional layout. Document Lens  also demonstrates space-based dynamics, where users can grab a rectangular lens and move moving it around on a grid documents to focus on the part of the document in detail where the rest is perspectively reflected onto the screen.
Object-based techniques consider each entity as an indivisible object, thus objects are the unit of operations. The WebBook and the Web Forager  employs object-based dynamics, where users operate on objects such as books, pages, etc. Graphical Fisheye View  is a distortion-based visualization technique that demonstrate object-based dynamics on the nodes of two-dimensional graphs. Size and location of each node is calculated in accordance to a distance to a focus point and an importance factor associated with each node.
Structure-based techniques take into account the structure of the objects. Structure determines which objects are affected and in which ways. The hyperbolic tree browser  provides an interactive visualization of large hierarchies of information represented in node-link diagrams, by employing structure-based distortion techniques. Treemaps  and the SHriMP  interface also demonstrate structure-based dynamics.