Singular value decomposition filter for speckle reduction in adaptive ultrasound imaging

Various adaptive beamforming methods have been developed recently for improvement of image quality in medical ultrasonic imaging. Such adaptive beamformers improve the spatial resolution and contrast, but often degrade the contrast-to-noise ratio (CNR) because speckles are well-resolved owing to improvement in the spatial resolution. For improvement of CNR, speckle reduction techniques are required. However, a typical smoothing filter, such as a Gaussian filter, blurs the boundary between different structures in an ultrasonic image. In the present study, a filter based on singular value decomposition was developed for improvement in CNR, while the sharpness of the boundary between different stucutures is preserved. The proposed method was evaluated by a basic experiment using a phantom with an anechoic cyst region. The CNR was improved from 2.93 to 11.06 dB by the proposed method.