The following images are a comparison of scan speed and filtering on laser scanning confocal images.

The original image was taken with 1 63X N.A. 1.4 lens at 1024 x 1024 pixels such that each pixel is 0.13 x 0.13 um. The image below is scaled to 25% and the cyan box represents the region that will be displayed at 1:1 pixels throughout the rest of the discussion of noise vs. speed and filtering. The images were saved as png instead of jpg to avoid spatial compression.

The speed setting in the Zeiss Zen software is the equivalent of the Hz setting in the Leica LAS software.

Upper left image was a single scan at speed 7. Upper right was each scanned line averaged 4 times. The result is less noise, but the scan time was 4X.
The two images below are the single scan speed 7 median filtered 1x1 and gaussian filtered 1x1 using ImageJ.

Note that some of the areas of noise in the fastest scan appear to have continuous structure in the other three images.

What about quantification of intensity?

Note that the mean intensities don't change but the average of four scans of real data and both filtering methods applied to the noisy data bring the standard deviation down by 2X or more.

The green area measured is relatively uniform. If we had filtered on small few pixel objects (a.k.a. high frequency), the intensity values would have significantly changed, but for broader areas (low frequency), note that the intensity values don't change significantly (the biological conclusion wouldn't chenge). While the speed 7 average 4 image is the best, perhaps the unaveraged image is sufficient. Your call.

So why not scan at 512 x 512 pixels at half the resolution and also less noise? You would lose fine structure, but if all you care about is intensity measurements, would be far more efficient.