Inorder to overcome the limitations of marker based AR, markerless AR was proposed. To find camera pose, natural features such as colour, shape, texture etc are used. The main objective is to track an object in each frame of video stream. Several methods of interest point detection are available for this purpose. Harris [3] is the most popular corner detector used in computer vision. SIFT (Scale Invariant Feature Transform) algorithm proposed by Lowe [6] is efficient to find interest points using Difference of Gaussian(DoG) of the image, and SIFT descriptor is based on histogram of gradients. SURF (Speed Up Robust Features) algorithm [2] is an approximation of SIFT in which DoG is approximated to box filters. Recently, several binary keypoint descriptors are proposed for efficient local feature matching in real-time applications.