| Compounding reduces speckle by averaging envelope detected data from several decorrelated looks. Directive filtering based compounding uses a Gabor-like filter bank to extract sub-bands from two dimensional RF data. Under the k-space response framework, each sub-band corresponds to using translated sub-apertures on transmit and receive. We extend the filter bank to allow sub-bands of different axial frequency centers, providing additional tradeoffs among lateral resolution, axial resolution, and compounding. Additionally, we propose an oriented demodulation that removes the 2-D carrier on each sub-band signal, which is necessary if the compounding averaging are to use complex baseband data. The methods were applied on in-vivo and phantom RF data for different filter bank configurations, yielding SNR improvements that are about 90\% of the best attainable increase with respect to the number of filters used. |