Canon released its 6D Mark II in 2017, and the Internet has been abuzz ever since then discussing its high ISO performance. There have been reports that the Mark II creates more noise at high ISOs than the original 6D.
But the thing about these comparisons is that, number one, it’s hard to two compare cameras precisely. Also, while it probably is possible to craft a perfectly controlled, laboratory-quality comparison to measure noise, it wouldn’t reflect real-life shooting conditions.
My step-dad, Jimmy, recently purchased a Mark II. Since I have the original 6D, we were curious to see whether the rumors were true. We set up two separate experiments to test our cameras in various shooting situations. While our experiment wasn’t laboratory quality, we planned it carefully and the results are surprising. Click here to skip ahead to the conclusion without reading about our process.
Background: What is Noise in Digital Photography?
My favorite definition of noise comes from Photography, by London, Stone, and Upton:
“Noise is the presence of pixels whose color and brightness is unrelated to the subject.” In other words, they are pixels that don’t belong in your shot – your camera’s sensor creates them, much like a weak signal creates static when you listen to the radio.
Learn more about noise here, including why you should overexpose to reduce noise.
You can see how these unrelated pixels affect a photo by measuring them with a histogram. Below is a photo of a color checker with a histogram to its right – I used the histogram to measure the row of neutral gray that you can see outlined. Each square on the color checker has one and only one brightness and color value.
The photo above was taken at ISO 100. Because the pixels contain minimal data that is unrelated to the scene, you can see how each column on the histogram relates to a square on the color checker. The columns are narrow and distinct.
Look at the difference in the histogram of the same color checker taken with an ISO of 25,600. In this photo, the histogram doesn’t have separate columns. That means that the brightness and color values of each square in the photo contain a range of colors and brightness levels that overlap from one square to the next. The pixels that fill in the spaces between where the columns should be are noise.
As if noise isn’t bad enough anyway, we actually have two types to keep track of: luminance and color noise.
Luminance noise is variation in the brightness of pixels that should have the same brightness. It looks like flecks of silver, gray, or white that shouldn’t appear in your photo.
Color noise is flecks of green, magenta, or other colors. It looks like this:
How to Measure Noise
Noise is hard to measure precisely in photos. For this experiment, I needed to measure luminance noise separately from color noise – since the Lab color space separates the brightness measurement from the color measurement, I used it to quantify the noise in our photos. Lab measurements consist of three values:
- L measures the luminance or brightness of a pixel
- a measures the pixels in terms of green and red. Negative a values are green, positive are red. 0 is neutral.
- b measures the pixels in terms of blue and yellow. Negative b values are blue, positive are yellow. 0 is neutral.
The a and b values of the Lab color space worked to measure the color noise. However, I didn’t expect the L values to be the same from camera to camera. Instead, to measure Luminance noise, I measured the width of the histogram for each color swatch, assuming that a more narrow histogram would have less noise.
The Process: Testing Canon 6D vs. 6D Mark II
The only variable was the camera.
Our controls were:
- The shooting settings (aperture, shutter speed, and ISO) were the same on each camera.
- Our lenses – each of us used the same model (the Canon 50 mm f/1.2) for one series of photos and the same exact lens (my 24-70 f/2.8) for another. We did the luminance and color noise tests using our 50s, and the “real life test” below used my 24-70.
- We used one stationary tripod – it didn’t move between photos.
- We took photos of exactly the same subjects.
By the way, can you tell that it’s science fair season in our family? Just imagine that I’m presenting these results on a neatly-laid-out trifold board…
We followed this process to compare the Canon 6D to the 6D II:
- We dialed in the same exposure settings on each camera, used Auto White Balance, and recorded our files as Raw.
- We set up the tripod and shot the scene with one camera.
- Without moving the tripod, we switched cameras and shot the scene with the other camera.
- To compare the photos, I set the Temp and Tint to the same measurements in Lightroom.
- I opened the photos as layers in Photoshop and used Auto-Align to make the pixels line up from one photo to the other. I used the luminance histogram and color selector tools to measure the pixels in our photos.
The Results: Canon 6D vs 6DII Noise Comparison
The photos were noticeably different from camera to camera. Each camera produced noisy shots at high ISO. While the appearance of the noise differed between the two cameras, neither of us could say that noise was better or worse from a cursory inspection of the photos. Because there was a difference, I wanted to quantify it to examine what caused it. This led me to inspect the measurements in Photoshop and revealed interesting trends.
Luminance Noise: 6D vs. 6DII
Because each photo was exposed slightly differently, I wasn’t able to measure the brightness numbers directly. Instead, I created a luminance histogram for the same part of each square on each photo and I measured the width of those histograms. Remember from my photo up above how photos with low noise have more narrow columns on the histogram? I used the luminance histogram to tell me how wide the range of darks and brights was for each square. The photos with the widest histograms have the most luminance noise. My histogram measurement process looked like this:
||Histogram Width 6DII
||Histogram Width 6D
|Medium Light Gray
|Medium Dark Gray
The 6DII had wider histograms, and more luminance noise, on 5 out of the 6 color swatches.
Color Noise Test: 6D vs. 6DII
To test the color noise, I used the Average Filter in Photoshop to get an average reading for the same pixels in each photo. These pixels should be neutral gray, meaning that both the a and b values of their Lab measurements should be 0.