The present invention relates to an interpolation method for enlarging an image, wherein the image is composed of plural scanning lines. Firstly, it selects four adjacent continuous pixels x.sub.k-1, x.sub.k, x.sub.k+1, and x.sub.k+2 which have the pixel values f(x.sub.k-1), f(x.sub.k), f(x.sub.k+1), and f(x.sub.k+2) respectively on each of the scanning lines; next, it determines three linear equations L.sub.k-1, L.sub.k, and L.sub.k+1 according to every two adjacent pixel values f(x.sub.k-1) and f(x.sub.k), f(x.sub.k) and f(x.sub.k+1), and f(x.sub.k+1) and f(x.sub.k+2), respectively; then, it selects a pixel x to be interpolated between the pixel x.sub.k and the pixel x.sub.k+1; finally, it applies the pixel x to the three linear equations L.sub.k-1, L.sub.k, and L.sub.k+1 to determine three candidate pixel values which are weighted and combined to obtain pixel value f(x) of the pixel x.