To improve the watermark benchmarks database, a new BRA (Block Replacement Attack) method is introduced in this paper. Unlike the typi-cal BRA which exploits the self-similarity among blocks in image, the scheme utilizes the self-simi-larity in various scale and orientation bands in the wavelet domain of the image. They search the re-placement blocks from the blocks of the smaller sub-image at the next lower resolution. The most important advantage of this scheme is that the pyr-amid sub-band decomposition acts as an automatic block classifier, with the result of simplifying the block search and block matching procedures and thus making the block replacement more effective. Furthermore, an improved algorithm based on Fuzzy set theory is introduced in which the num-ber of blocks chosen for combination are modified adaptively to obtain the candidate replacement blocks. It simplifies the procedure for searching the candidate replacement blocks. These advantages are confirmed by the experimental results, whichshow that the performance of the scheme is supe-rior to typical BRA methods.