The efficient motion estimation algorithms are the important research hotspot in the real-time video coding-decoding technology. In order to reduce the high computational complexity of motion estimation in the video-coding, a fast adaptive motion estimation algorithm based on the context is proposed. The novel algorithm uses a kind of parallelogram pattern to search directionally, and imports secondary minimum distortion point. The basic idea of the novel algorithm is: In the process of block-matching, this novel fast algorithm uses the former search step's motion vector and the relative location between the minimum distortion point and secondary minimum point to select adaptively one parallelogram model and decide the search direction of next step, and then keeps on searching for more accurate motion vector. Experimental results show that the proposed algorithm reduced the computational complexity of the motion estimation in video-coding effectively.