I assume you set ISO 2500 because you wanted a nice looking jpeg sooc. Unfortunately the walls of the building are still very noisy which is a shame given the very nice scene and composition you chose.
Fwiw to anyone interested, to minimise visible noise, especially in low light scenes, you need to get as much light as possible onto the sensor within your DOF and blur requirements. It is a low amount of light onto the sensor which results in a low signal to noise ratio in the raw data which in turn makes the noise more visible.
In a scene like this where there is no movement, assuming the trees weren't swaying too much in a breeze, and you are using a tripod the best way I know to maximise the amount of light onto the sensor is to ETTR.
-
Set the widest aperture (smallest f-number) that will give the DOF you want.
-
Set base ISO
-
Use the camera's live-view with a histogram and slow the shutter speed until important highlights just begin to clip - ie...push the histogram as close as possible to the right hand side until just before important highlights begin to clip.
This will maximise the amount of light onto the sensor within your dof and blur requirements which will result in maximising the SNR within those requirements. This will result in the minimum amount of visible noise after you then dial back the image lightness to what you want either in your raw converter, preferably, or in your sooc jpeg image editor. The benefits of ETTR are greater when using raw files (16 bits) rather than sooc jpegs (8 bits).
Your denoising app will thank you for making it much easier to clean up an ETTR'd raw file than a non ETTR'd jpeg.
Assuming base ISO on your camera is 100 then had you set base iso instead of ISO 2500 and ETTR'd, you would have been able to put at least an extra 4.66 stops of light onto the sensor. That's a huge amount of extra light and would have reduced the amount of visible noise very significantly.
If the display on the outdoor screen was changing frequently I would just simply take a few separate shots of the screen displays (exposed appropriately) and then in post insert the one that works best into the final image.
Anyway, just some food for thought to maybe consider in similar situations in the future.
