Sonar images are typically affected by a granular pattern interference known as speckle noise, which degrades image contrast. To aid in object detection and recognition for speckled imagery, a robust version of the Lee filter is presented. The new method essentially combines robust statistics with an adaptive approach to achieve an effective balance between contrast stretching and speckle reduction. Tests were performed on real sonar images, where objective metrics and direct visual perception were used to evaluate the results. Experiments have shown that this easy-to-implement filter remarkably highlights edges and details with apparent speckle reduction, offering a promising simple tool that may be useful in segmentation and classification applications.