This is generally directed to systems and methods for noise reduction in high dynamic range (“HDR”) imaging systems. In some embodiments, multiple images of the same scene can be captured, where each of the images is exposed for a different amount of time. An HDR image may be created by suitably combining the images. However, the signal-to-noise ratio (“SNR”) curve of the resulting HDR image can have discontinuities in sections of the SNR curve corresponding to shifts between different exposure times. Accordingly, in some embodiments, a noise model for the HDR image can be created that takes into account these discontinuities in the SNR curve. For example, a noise model can be created that smoothes the discontinuities of the SNR curve into a continuous function. This noise model may then be used with a Bayer Filter or any other suitable noise filter to remove noise from the HDR image.