Pixels Redux

It's interesting how one topic so naturally leads to another in photography. We're juggling so many decisions and issues, and they are all interrelated.

In answering one person's question about yesterday's article, we veered into one of those relationships. When I was writing about pixels yesterday I was just assuming "perfect pixels." In other words I was writing about, oh, 12 million perfect pixels versus 24 million. In other words, I was in the abstract (it was an article focused on an abstract notion, after all: forest or tree?).

In the real world, we get real pixels, not perfect ones. And here's where things get interesting. While sensors have certainly improved in the 12 years of the Nikon DSLR era, something else has actually improved pixel quality more. I'm referring to the imaging ASIC, or what Nikon calls EXPEED these days.

Go back and grab a D1h/D1x and use it on a JPEG setting. Dial up some sharpness and JPEG compression. Oops. Nasty things start happening at the pixel level. Now try that with a D3s. Heck, I can dial up as much compression as the camera allows and barely see a difference. It isn't the sensor that's doing that (though starting with a less noisy signal is helpful), it's the JPEG processing engine.

I see this all the time in my testing. I've had several competitive models through my lab that used the same sensor but have very different looks and results. Let's take m4/3, for example. From the same sensor, the Olympus E-P1/2/PL-1/2 and the Panasonic G/GF-1/2/et.al get very different JPEG results. From the raw data out of these cameras, I can get nearly identical results, which is why I'll take the camera that gives me the right set of features and other performance. The difference in JPEG? True Pic (Olympus) versus Venus (Panasonic). Olympus has Panasonic beat, hands down. In maturity of imaging ASIC and its ability to create excellent JPEG images from underlying raw data, I'd rank the DSLRs this way:

  • Canon Digic. Say what you want, it's mature, creates impressive final images from even questionable data, and has a bit of contrast tweaking and edge sharpening missing in Nikon's.
  • Nikon EXPEED. Not far behind, but Nikon has chosen to go for smooth data rather than abrupt edges for some time now, resulting in what people say are slightly soft JPEGs. But they withstand added sharpening and don't have that hint of falseness you see in some other imaging ASICs.
  • Toss-up: Sony Bionz, Olympus True Pic. Olympus might get the nod here because they're working with crummier data, and Sony is often using the same data as Nikon and getting lesser results. That speaks to the maturity of the Olympus routines.
  • Also-rans: Panasonic Venus, Samsung ???. Both need some work on improving their JPEGs. Panasonic clearly isn't getting what Olympus is from the same sensor, and Samsung's results, while good, have some minor issues in them, too.
  • Can't Rate: Pentax Prime. I don't have enough experience with their DSLRs to get a good handle on where they stand. What little experience I've had would put them somewhere at or near Nikon in this list, I think. But I just haven't used enough Pentax DSLRs, especially the latest ones, to judge.

For JPEG capture, the imaging ASIC makes a huge difference. That's because JPEG is actually throwing away enormous amounts of pixel data. We reduce from 12- or 14-bit to 8-bit, for example. We further reduce as much as 100MBs of RGB pixel data down to 6MBs of formula. If you don't think there are better and worse ways of doing all that data reduction, well, I have some land alongside the Mississippi River I'd be happy to sell you right now.

When we talk about real, post JPEG processing pixels, we're no longer talking about perfect pixels. JPEG engines can (and most do, though not so much Nikon) artifact, they can apply noise reduction that smudges acuity, they can alter contrast curves in ways that boost one area of contrast versus another (usually mid-range), they can halo on edges (though not generally Nikon), and much more.

If you're using JPEG, getting "sand" instead of "shape of the coastline" is going to be more difficult, and on some cameras, near impossible.

Looking for gear-specific information? Check out our other Web sites:
Nikon Z System mirrorless: zsystemuser.com | Nikon DSLRs: dslrbodies.com | Nikon film SLRs: filmbodies.com
Privacy Policy | Sitemap

bythom.com: all text and original images © 2026 Thom Hogan
portions Copyright 1999-2025 Thom Hogan