Medical ultrasound is an ultrasound wave-based imaging technique used to scan internal structures of the body and is non-invasive, portable, accurate, and cost-effective. Thanks to these characteristics, this imaging modality has gained great importance in the diagnostic field of radiology. However, a major limitation of ultrasound images is the presence of locally correlated multiplicative noise that degrades image quality. Speckled noise is generated by interfering echoes of a waveform transmitted at the time of acquisition. The superposition of approaching sound echoes with random time periods and amplitudes produces a complicated interference pattern called speckle noise [1]. Speckled noise is likely to create uncertainty and may hide vital clinical facts, thus skewing the opinion of radiologists. Therefore, a despeckling technique is needed, which can restore an image without altering the important features and structure of the original image so as to assist radiologists in making a more accurate diagnostic decision. The main objective of this work is to develop a versatile despeckling technique to improve the quality of medical ultrasound images for better clinical diagnosis. The proposed technique is a variant of the technique recently published by Sudha et al. [2]. The innovative aspects of the proposed technique are threefold: (a) use of curvelets to efficiently separate edges and noise, (b) revised criteria to define the weights of the window model used to calculate the weighted variance by exploiting the intra-dependencies of band present in the curve coefficients and (c) devised a window tuner to help radiologists control the degree of smoothing of the output image. The document is organized as follows. A short...... half of the paper......ne in curvelets rather than denoising using wavelets [28]. The hard threshold was used to modify the curve coefficients [28-32]. The literature survey reveals that very little work has been reported for despeckling using curvelets [29]-[32]. None of these techniques [29]-[32] take into account the intra-band directional dependencies of the curvelet coefficients for the threshold calculation, which are very important to make the threshold adaptive to the signal variation. Dot removal is always a trade-off between noise suppression and information loss, something experts are very concerned about. It is therefore interesting to keep as much important information as possible. In this paper, an attempt has been made to design a scaled adaptive threshold despetting technique in the curve domain to recover an image without losing the vital fine curved details.