Method of decoding images and device using same

Scalable video encoding uses interlayer texture prediction, interlayer motion information prediction, and interlayer residual signal prediction in order to remove redundancy from interlayer images. In order to increase the accuracy in interlayer prediction, the present invention may find a reference layer block on a location corresponding to the current target block and a block that is most similar to a sample of the current target block from images of a reference layer and use them as a prediction signal. Also, in interlayer prediction, a prediction signal obtained from an intra-layer image to which the current target block belongs and a prediction signal obtained from a reference layer image may be weighted and then used as a prediction signal.