Abstract: A novel wavelet image coder based on morphological dilation and run-length coding, dilation-run algorithm, is proposed according to both intra-subband clustering of significant wavelet coefficients and inter-subband dependency in significant fields. The algorithm mainly involves discrete wavelet transform, bit-plane coding, morphological dilation of significant coefficient cluster within subbands, coefficients significance predicting across subbands, run-length coding of the insignificant coefficients before each seed and adaptive arithmetic coding. The two key features of the algorithm are morphological dilation and run-length coding. The morphological dilation operator is used to search and encode the significant coefficients in each cluster, while a new run-length coding method is used to encode the position of the seed of each cluster, which is the start pixel of the dilation. The dilation-run algorithm, as a kind of embedded coder, is also rate scalable. The experimental results show that the new coder outperforms the zerotree coder SP1HT and is competitive with the morphology coder MRWD and SLCCA. For images with strong clustering feature after the wavelet transform, the new coder outperforms both the morphology coders above.