A new effective feature extraction scheme, modifying SURF(Speed Up Robust Feature) algorithm for markless AR, is presented in this paper. SURF algorithm used the scale pyramid for invariant feature and the block based NMS of a fixed size window, so that the algorithm requires a lot of calculations. The proposed algorithm suggests an enhanced corner detection based on edge instead of scale pyramid, in order to assure invariant features, and uses an adaptive NMS(Non-Maximum Suppression) in the interested area for more effective calculations. As the result, the proposed algorithm preserves the performance of feature extraction similar to that of the conventional SURF, but obtains superior results in computation loads and time.