Adaptive pyramid filtering for medical ultrasound image enhancement

This paper describes an adaptive pyramid filtering method for increasing medical ultrasound image qualities. Most medical ultrasound scanners employ logarithmic function to reduce dynamic range of the echo envelop signals. In our method, a log-compressed speckle image is first decomposed into multi resolution representations using the Laplacian pyramid (LP). Each LP layer is then filtered with an adaptive filter to smooth noise and save features at the same time. The controlling parameter of the filter at each LP layer is calculated using the local statistics and the transformation function of the equivalent filters related to the LP layer. The output image is then reconstructed from the filtered LP layers, with speckle reduced and edge preserved effectively. Experiment results using a phantom data and in vivo data show the efficiency of the method