Image enlargement techniques aim to convert low-resolution images into high-resolution ones. Conventional linear interpolation methods usually have difficulty in preserving local geometric structures. Nonlinear interpolation methods could well exploit the local data properties and thus obtain better results. However, their computational complexities are much higher than those of the linear methods. In this work, we classify an image point into four classes of texture constitutions by analyzing the local properties of its derivatives. We then adaptively choose appropriate interpolation scheme according to the constitution of a pixel. Experimental results demonstrate that the proposed method produces higher visual quality than conventional methods. Furthermore, it is suitable for hardware implementation due to its low complexity.