This paper proposes a novel Adaptive Rood Pattern Search (ARPS) algorithm for block matching in video shot boundary detection. ARPS uses motion activity, which is measured by the magnitude of motion vectors between frames, to help predict the motion vector for each block. The algorithm works by first assuming blocks near a current block will move in a similar direction, then searching in a pattern around the predicted motion vector. Experimental results on test videos show ARPS can accurately detect shot boundaries with fewer false detections than other block matching algorithms.