We perform a statistical analysis of curvelet coefficients, distinguishing between two classes of coefficients: those that contain a significant noise-free component, which we call the “signal of interest” and those that are primarily noise. This distinction is based on the fact that the noise-free component of the coefficients is typically more concentrated in the lower frequency bands, while the noise is more spread out. By modeling the distribution of the noise-free component, we can effectively remove the noise from the coefficients, leading to improved image denoising performance. The proposed method is evaluated on a variety of images and compared to state-of-the-art denoising techniques, demonstrating its effectiveness in removing noise while preserving image details.