Live View is a terrific feature.  For landscape photographers, there are two benefits I especially like.  One is critical focus ... I can pop into Live View and then 10x magnification and manually focus, which can help insure my focus is accurate.  After all, auto-focus isn’t completely reliable.  The other is a mirror lockup function ... easier than popping into custom settings and enabling that.  (I’m talking about Live View as it relates to Canon bodies here.  I don’t know how it works on other brands).

Any time the sensor is energized it can create heat, and heat can create noise.  It stands to reason excessive Live View can add noise to your image, offsetting any quality gained by the use of mirror lockup and the exacting focus it offers.

I’m not sure of a way to test this quantifiably.  I know Canon claims it doesn’t, but I would suspect they have a “tolerance”, so some possible noise may be acceptable to them.  That tolerance may be OK for most, but anyone using Live View for the purposes I mentioned is trying to exact the ultimate image quality, so any measurable increase in noise is probably unacceptable, especially since there are alternatives to using Live View to accomplish both functions.

But it’s so convenient ... it would really be nice if using Live View for these didn’t add any noise.  I decided to try a test of sorts.  It’s not scientific, and really isn’t “conclusive”.  But to me it indicates that if Live View is adding noise, it is probably insignificant.

The idea I had was pretty simple.  First I would capture 6 “black” frames.  The first one would determine a baseline, then Live View would be activated and each minute for 5 minutes another frame would be captured.  The body cap was left on, and the Viewfinder covered so light couldn’t enter the camera.  I did the test with the new 5D Mark 2.  Exposure was for 1 second at ISO 100.  I realize this isn’t a “real world” comparison, because normally in Live View light is hitting the sensor.  This may affect the sensor, but it seems if this increased noise it would be difficult to create video, which the 5D Mark II does very well.  All in camera noise reduction was disabled, including Long Exposure NR.

The exposures were then brought into Lightroom, where noise reduction and sharpening was disabled.  The black level was set to 0, and the exposure value was raised to ev +4.  This leaves us with a pretty noisy image.  At that point I opened them in Photoshop CS4 and did a screen capture of each exposure’s histogram.  I then stacked them into a single document, selected a 250 pixel square, and cropped the document.  This insures each image is from the same pixels on the sensor.  Here are the results...

I’ll be the first to admit this isn’t conclusive, but it sure appears that after 5 minutes of Live View their isn’t much difference in the image or the histogram.  The histogram shows the mean for the baseline exposure at 14.42 and the five minute exposure is 14.41.  All of them show the median at 12, and the standard deviation for 5 minutes is 9.71 vs 9.92 for the baseline - which implies that the 5 minute exposure may have less noise.

I’ll probably play with this a little more, and maybe even research the meaning of the information to try and make sure I’m understanding those numbers right.  But for now I’ll continue to use Live View ... it doesn’t appear to affect the noise of the capture at all.