Oscillatory textures are not only ubiquitous in natural images, but also in some inverse imaging problems involving oscillatory date, such as seismic databases, fingerprint images, et al. In this paper, we propose a three-step denoising scheme for for oscillatory textural images by combining bilateral filtering in spacial domain with wave atoms shrinkage method in transformed domain. That is, we first pre-process noisy image with bilateral filtering, then to process image with wave atoms shrinkage, finally post-process image with bilateral filtering again. Above all, we offer a reasonable interpretation of proposed method from the point of nonlinear diffusion filtering. Numerical experiments illustrate the good performance in comparison to the wave atoms shrinkage method and the bilateral filtering method by using two measures: peak signal-to-noise ratio (PSNR) and structural similarity (SSIM).