We propose a non-blind digital image watermarking method that combines the Dyadic Wavelet Transform (DYWT) and the Gyrator Transform to enhance watermark robustness and imperceptibility. The method involves decomposing the original image using the DYWT and applying the Gyrator Transform to the low-frequency component. We precisely embed the watermark into this region by predicting the low-frequency region after transformation using a nonlinear regression model, increasing its resilience to various attacks. Experimental results demonstrate that the proposed method achieves high image quality, with Peak Signal-to-Noise Ratio (PSNR) values exceeding 35 dB and Structural Similarity Index Measure (SSIM) values above 0.98. The method is robust against common image processing attacks such as filtering, JPEG compression, and geometric distortions like cropping and rotation, outperforming existing methods. The Normalized Correlation (NC) values remain high after various attacks, indicating reliable watermark extraction. These results confirm the effectiveness of the proposed non-blind method in achieving both high imperceptibility and robustness in digital image watermarking.