Interpolation of Medical Ultrasound Images From Coherent and Noncoherent Signals

Medical ultrasound imaging instrumentation typically performs image interpolation using the signals acquired after envelope extraction, i.e., noncoherent signals. This operation is completely satisfying when the Nyquist condition for spatial sampling is fulfilled. However, the maximum spatial frequency of a signal increases as a consequence of the envelope extraction, making it more difficult to fulfill the Nyquist condition. For this reason, this paper suggests the interpolation of signals before envelope extraction, i.e., the use of coherent data, and discusses the applicability of this procedure to actual ultrasound scanners. Emphasis is given to the linear-array imaging technique, and the adoption of windowed sinc functions as interpolation kernels with limited support is assessed. Moreover, the authors investigated the situation in which coherent signals are also undersampled, and perfect reconstruction is not at all possible. The investigation was carried out using a numerical study and a simulated imaging test, and it was supported by subjective and objective evaluations. It is concluded that, at a moderate undersampling level, the interpolation of coherent signals allows significantly better performance than the interpolation of noncoherent signals.