Iteration in conventional fractal coding schemes not only leads to a high computation complexity, but also requires a large amount of memory. It unavoidly prolongs the decoding process and precludes the use of high speed applications. To overcome these problems, we propose a non-iterative method based on a novel domain pool design for fractal image coding. The domain pool used in the encoder is on-line transmitted to the decoder. Therefore, the domain blocks are generated from the same mean image existing in both the encoder and decoder. We next ultilize contractive affine transformations to encode/decode the image with iterations. From the simulation results, we have successfully speeded up the decoding process, and the coding performance for the test images is good as or even better than that of the conventional schemes.