A digital image watermark embedding and extracting algorithm is presented based on the multiscale ridgelet transform (RT) which can efficiently represent image with linear singularities.The RT transform also has directional sensitivity so that among the transformed coefficients the most significant ones represent the most energetic directions of straight edges in an image.To get straight edges the digital image with curvilinear edges or texture is partitioned into small pieces.Then the pieces are treated by RT and the most energetic directions of pieces whose energy are greater than some threshold are selected.And eventually the watermarks are embedded in the the middle ridgelet subband of the selected directions.Different noises with different SNR are added to the watermarked image in the experiments and the results prove robustness and transparency.