Abstract: In the ultrasound color flow imaging (CFI), a clutter rejection method based on the projection-pursuit (PP) is proposed to reduce velocity artifacts caused by clutter signals. According to time domain characteristics of echoes, principal components are extracted by the PP to reduce their sensitivity to strong clutter signals originating from non- blood-flow regions. Blood signals are then obtained by removing the estimated clutter space in the eigenvector domain. The proposed method has been applied to both computer simulated data and human experimental data. Results show that this method can achieve a better velocity profile and vessel shape than the conventional high-pass filter. It is also shown that this method causes less velocity artifacts in non-blood-flow regions and achieves an inside to outside blood energy ratio higher than that achieved by eigenvector method by 5 dB. Thus, the flow image quality is expected to be improved effectively.