Abstract This paper describes a digital image processor for ultrasonic speckle suppression that was explicitly designed to satisfy the requirements of detail preservation, adequate smoothing and real-time operation. The first two of these requirements were addressed by selecting a nonlinear adaptive algorithm, which uses a measure of local homogeneity to adjust the amount of smoothing performed at each point of the scan, and employing a large (9 × 9 pixels) filtering window. Real-time operation was achieved by developing a highly concurrent systolic architecture that allowed the efficient mapping of the algorithm into low-cost high-density hardware. Initial experience, obtained by interfacing the processor to a scanner, suggests that this type of processing is capable of enhancing the visibility of subtle differences in echogenicity while retaining genuine image detail, as judged by the preservation of the small vessels and ducts. In addition to its impact on contrast resolution, this form of real-time speckle suppression should also prove valuable as a preprocessing stage before performing other digital image processing operations that are sensitive to noise, such as segmentation and three-dimensional rendering.