Abstract—Fiducial images are a common method for supporting vision-based tracking in augmented reality systems. This paper addresses the question: what is the best fiducial? A set of criteria that are desirable in an optically tracked fiducial are presented and a new fiducial image set is designed that meets these criteria. The images in this set utilize a square black- border pattern with a 15% border width and an interior image that supports orientation determination and unique identification. The interior image is constructed from orthogonal DCT basis images chosen to minimize the probability of misidentification and to be robust to noise and occlusion. We describe how this image can be integrated into an AR software system such as ARToolKit. Keywords: Augmented reality, fiducials, motion tracking I. INTRODUCTION Fiducial images, often referred to as markers, are a popular element of many of the vision-based tracking systems utilized in augmented reality applications, including the ubiquitous ARToolKit. However, images used as fiducials are often selected on an ad hoc basis. This paper addresses the question: what is the best fiducial? Should a fiducial image be round, square, or even triangular? What are the best colors to utilize? Should the image be a bar code or some other blocky pattern? Clearly, there are tradeoffs, so a set of criteria are specified in the paper that make it easier to answer that question for a wide variety of applications. Fiducial images are images placed in a physical environment in support of tracking, alignment, and identification. Train cars have bar codes that allow machinery to automatically identify and route them through stations. Circuit boards have fiducials that allow masks to be aligned from layer to layer and allow the position of the board in a jig to be precisely measured so that robotic machinery can properly insert components. In augmented reality systems, fiducials are generally used for tracking elements in the environment. They may be placed in the fixed, physical environment so that the location of a moving camera can be identified or they may be placed on moving objects or people so that a location relative to either a fixed or moving camera can be computed. AR systems commonly rely on tracking to determine head position and orientation in support of rendering graphics registered with the surrounding environment. Proposed fiducials have been as simple as small dots in a pattern or as complex as bar-coded circular or square images. The most relevant examples to this conference are the ARToolKit markers, square fiducial images with a fixed, black band exterior surrounding a unique image interior [1]. Figure 1 is an example ARToolKit fiducial. The outer black band allows for location of a candidate fiducial in a captured image and the interior image allows for identification of the candidate from a set of expected images. The four corners of the located fiducial allow for the unambiguous determination of the position and orientation of the fiducial relative to a calibrated camera. Figure 1- Example ARToolKit Fiducial Additional examples of fiducial images include the TRIP (Target Recognition using Image Processing) system, the nested multi-resolution colored ring system, and CyberCode. TRIP is a vision-based sensor using visual markers in the form of rings [2]. One ring is used as an ID code with a very large space. The other rings provide for synchronization information and support for the POSE_FROM_CIRCLE algorithm [3]. Cho, Lee, and Neumann utilize nested colored rings for fiducial images [4]. The nesting allows the rings to work over a wide range. CyberCode is a bit-based fiducial similar to a 2D bar code [5]. Many simple approaches using fixed color squares, circles, or cross patterns have been demonstrated. Most projects approach the problem either from the standpoint of selecting a set of images (as in ARToolKit) or choosing a way to encode data into images (as in CyberCode). Charles B. Owen, Fan Xiao, Paul Middlin Media and Entertainment Technologies Lab Media Interface and Network Design Lab Michigan State University 3115 Engineering Building East Lansing, MI 48824 cbowen@cse.msu.edu What is the best fiducial?