The integration of virtual and real environments promises to yield many interesting applications that combine the natural, tangible and rich interactiveness of physical reality with the freedom from constraints and precision of the digital world. Wearable computers are a particularly convenient platform for this symbiosis and make this integration as seamless as possible [17] [12]. One area where this cooperation would be especially interesting is in the game of billiards. This pastime involves a complex blend of physical dexterity as well as analytic geometry and strategy. To enhance the experience, it would be desirable to supplement the user's ability to select, visualize and aim without reducing the tactile feedback of the real game [8]. This intermediate form of integration lies conceptually somewhere between a merely physical game of billiards and a sterile and unnatural computerized version of it. A middle ground between these two extremes can be achieved by using computer vision to map real world data into a virtual context. Similarly, a head mounted display could remap the virtual results of any strategic analysis onto an image of the user's surroundings [6]. The use of a wearable computer allows the user to navigate freely in an arbitrary physical space without special preparation.
This paper describes a wearable computer-based system that uses
probabilistic computer vision to guide the user's game play throughout
a match. We expect the user to be a novice player who has yet to
become familiar with the game and can quickly benefit from the system
as a training aid. Figure 
depicts the desired output of the
system. Given an image of a pool table, compute the best shot by
locating the pockets and the balls and computing the easiest shot that
can be performed. Then, display the paths the relevant balls should
follow.
Although we have chosen to focus on one particular application (billiards enhancement and training), the motivation behind this work is the integration of augmented reality with computer vision and wearable devices. This integration brings a new dimension to computer-human interaction by essentially reducing the interface barrier. The computer is brought closer to the user in physical terms (wearable), in input (vision and perceptual intelligence) and output (augmented reality and visually integrated output).
Figure: The Desired System Output
We begin by describing some earlier work in the areas of wearable augmented reality and entertainment. We then introduce our system, describing the hardware equipment and the required algorithms. The vision algorithm components including probabilistic color modeling and symmetry computations are explained. The classification and recognition of different balls and pockets is discussed. Subsequently, we present a method for ranking strategic shots from the computed loci of the balls and pockets. The integrated system is finally evaluated and important issues as well as future work are presented.