What is the most honest way to cite pixel count (resolution)?

Question

Canon cites the following features for the EOS-1D X:

Total Pixels: 19.3 Megapixels
Effective Pixels: 18.1 Megapixels
Largest Image: 17.9 Megapixels (5184 x 3456)

The discrepancies are explained very well here: Why are effective pixels greater than the actual resolution?

I am writing an objective comparison of different cameras and need a standard way to compare resolutions. The comparison is targeted at end users. I believe the resolution of the largest image (ie. 17.9 Megapixels for the EOX-1D X) would provide the most practical and fair representation of a camera's resolution to use as a basis for comparison. Is this correct and if so/not, then please explain in technical terms why so/why not.

Answer 1

I would use the pixel count from the largest image, as that is what we are actually seeing when we view our images, RAW or otherwise, on a screen.

The effective pixels is the actual sensor pixel count, which is usually a little larger than the image pixel count due to the way interpolation works (you linked to a topic where I provided an animated image that demonstrates why already.)

The total pixels is larger than the effective or image pixels because most sensors included a border of masked off and black pixels. These pixels are used for calibration purposes, and are included in the RAW, but are not processed during debayering/interpolation into the final on-screen image. This pixel count may be important in knowing why a RAW image file is as large as it is, because these extra pixels ARE usually included in the RAW (along with any metadata). It may also be important to know the total pixel count when trying to determine the data throughput rate of a camera. For example, most people computed the 1D X data throughput rate incorrectly at around 440mb/s (~222mb/s per DIGIC 5+) by using the 18.1mp pixel count. I computed the data throughput rate at 500mb/s (250mb/s per DIGIC 5+) by using the 19.3mp pixel count and factoring in overhead for metadata.

Outside of these fringe cases, the only pixel count most people are really going to care about is the image pixel count.

Answer 2

Yes, the resolution of the largest image makes sense.

However, really, these days, this is barely an important characteristic — all cameras on the market are at the "yep, that's a lot" level or beyond. It's true that more megapixels can be good, but in order to take advantage of them, one needs a very good lens, a lot of light, no camera motion, and so on — for most use, the difference between 16MP and 20MP is negligible. (As a rule of thumb, I'd suggest that 50% larger in each dimension is a significant difference — so, more than double the megapixel count.)

You could go with the "perceptual megapixels" score introduced by DxOMark — this tries to give a meaningful number for the practically perceived detail for a camera and lens complication. But to me, this is a lot of extra complication, and overall I'd just deemphasize the number overall — maybe lumping cameras into 10mpix-ish, 20mpix-ish, 40mpix-ish, and so on.

To be really helpful to "end users" (I'm not sure exactly what that is in a photographic context, but I'll take it to mean beginner/intermediate photographers), I might even suggest grouping the cameras into high, medium, and low pixel count with a * noting that the designation is relative to other cameras of the same approximate age and with the same sensor size. For today's mid-range cameras, that might be

• Low: < 12mpix
• Medium: 12-20mpix
• High: 20mpix+

but you might want to use different scales for camera phones (with 8-12mpix being medium) or large-sensor full-frame cameras (with medium at 24mpix, say). In any case, the key is to use a generious range — don't break it down more than is meaningful.