Methods The proposed approach combines 2D spatial wavelet filtering with 1D temporal Karhunen-Loeve Transform (KLT). The KLT is first applied to create a series of “eigenimages” in which important signal information is concentrated into only a few eigenimages. Then a 2D spatial wavelet filter is applied to each of the individual eigenimages. An adaptive threshold is used to define the wavelet filter strength for each of the eigenimages based on the noise variance and standard deviation of the signal, resulting in stronger filtering of the eigenimages that primarily contain noise. After wavelet filtering, the de-noised eigenimages are transformed back into image space. The performance of this filtering approach was tested over a range of wavelet filter strengths (1x to 4x adaptive threshold) for noise reduction and edge sharpness in a digital phantom. SNR improvement was also evaluated in real-time stress cine image series acquired in five normal volunteers. SNR was calculated in phantom and human images by finding the mean signal value and noise variance in a region-of-interest (ROI) within each frame. The sharpness in phantom images was measured as the distance between 20% and 80% of the total rise/fall for a moving edge.