How to select the finite ridgelet transform (FRIT) coefficients which were suitable to embed watermark data was an interesting problem. Digital watermarking algorithms in FRIT domain were accomplished by modulating the transform coefficients, which represented the most energetic direction of an image. However, these coefficients which were the most suitable for embedding watermark were short of study. Based on Laplacian probability distribution function (PDF), the FRIT coefficients which were particularly important for the perceptual effect of the embedded watermarks were analyzed by adopting maximum a posterior (MAP) estimate. The embedded watermark could be detected blindly by correlation coefficients between the watermarked image and the watermark sequence. Experimental results prove the robustness and transparency.