To solve the problem of blurry edge in traditional linear interpolation algorithm, an improved algorithm is proposed. Firstly, we compute six interpolations of horizontal, vertical and diagonal direction with inverse distance square approaches in the neighborhood of the interpolation point, and then we define weighting factors with interpolation distance and direction gradient. Finally, we get the final interpolation by data fusion. Considering both the distance of interpolation and the direction gradient, the information of the edge and the texture of the original image are both effectively protected in this method. Results show that compared with the traditional bilinear interpolation approaches, the new one makes the mean squared error decrease, while the average gradient increases, which is effective to improve image resolution.