In this paper, we present a compressed-domain watermarking technique for still images based on the SPIHT (set partitioning in hierarchical trees) coding. The proposed watermark is engraved on the refinement bits of SPIHT. By jointly considering watermarking and compression, the proposed quantization-based watermark suffers less degradation from the host interference and lossy compression. Watermark verification can be realized through a low-complexity hard-decision detector without host information. We further present a repetition-code scheme to enhance the robustness performance of the system. Experimental results show that the proposed method is more resilient to non-geometric attacks than the transform-domain spread-spectrum methods.