It has been estimated that one out of every four medical diagnosis in the world involves ultrasound imaging modality because of its noninvasive nature, low cost and capability of forming real time imaging. Ultrasonic imaging extends its application to many fields of medical diagnosis, but the utilization is being unfortunately affected by speckle noise. In this article, an efficient multiscale approach is proposed to reduce speckle, to enhance the edge information and to preserve point and linear features, rather than just inhibiting smoothing. With this approach, the image enhancement is made in three steps: First the image is transformed into Laplacian pyramid domain representation. Second, the pyramid coefficients are manipulated by permutated diffusion, and finally the image is reconstructed from the diffused Laplacian pyramid. New permutated diffusion is proposed for coefficient manipulation for effective speckle reduction and enhancement. The proposed permutated diffusion avoids the blocky effects caused by second-order partial differential equation (PDE) and requires only little iteration compared to fourth-order PDE to converge. In each pyramid layer, a gradient threshold is estimated automatically using robust median estimator. The mean absolute error between two adjacent diffusion steps is used as a stopping criterion. Performance of the proposed approach is compared with the state of the art pyramid based methods. Experiments on synthetic data, simulated phantom and real ultrasound data set indicate effective suppression of speckle, preservation of edge information and their structural details.