OMD with more MP?

Started 2 months ago | Questions thread
Anders W
Forum ProPosts: 17,051Gear list
Re: all your megapixels are fake
In reply to Great Bustard, 2 months ago

Great Bustard wrote:

Anders W wrote:

Great Bustard wrote:

Alex Notpro wrote:

Are you fully considering "information content" of each pixel? Is there an additional layer of information interpolation lurking in there?

There's spatial information and color information. From what I've seen, the 2x approximation for Bayer vs Foveon seems about right. For example, that means that the current 15 MP Foveon sensors should resolve as well, on average, as current 36 MP Bayer sensors, lens permitting, of course.

Should perhaps, but do they? Here are a set of crops from (aka Lenstip when translated to English) showing the 36 MP Bayer D800 versus all three 15 MP Foveon Sigma DP Merrills. The images were processed (by from RAW by means of DCRAW (no sharpening, no NR). The crops from the Merrills were upsampled (by me) in PSE 2.0 to roughly the same size as that from the D800. Does it really look like even-steven resolution-wise to you?

Note: Click on the image to view it properly.

This is a very interesting comparison. There is a significant amount of detail in the D800 photo absent in the DPM photos, which is very curious.

Yes. After a discussion on this forum about the prospect of an MFT camera with a Foveon sensor a while ago, I have been taking a rather careful look at the usual claim that an X MP Foveon is equivalent to 2X MP Bayer sensor in terms of resolution and found that it doesn't hold water. In practice, Bayer does better than that. I think the reasons are essentially twofold.

First, the loss due to Bayer sampling is in reality less than one might expect based on just counting the number of green pixels versus total pixels thanks to smart demosaicing algorithms. This is what shows up in the comparison above.

Second, one serious weakness of Foveon sensors is that the color separation isn't very good. In order for the final colors to be as saturated as those from a Bayer sensor, you need to crank up saturation significantly (i.e., amplify the weak color differences). When you do, you get significant amounts of chroma noise, not only at higher ISOs but at base ISO too. Look at the samples I posted in my reply to Alex Notpro here (which are all at base ISO, which I forgot to mention)

and you'll see what I mean. To get rid of that, you need to apply some chroma NR, desaturate a bit (especially in the shadows), and/or "crush" the shadows (letting them quickly go close to black). The Sigma RAW converter and OOC jpeg engines use all three strategies. But no matter how you handle it, much of the theoretical advantage of Foveon for color detail is effectively lost.

While Foveon samples all three colors rather than just one at every pixel, the sampling is so noisy that it doesn't possess much of an advantage in comparison with the Bayer strategy. The way the new sensor for the Quattro is designed is an implicit recognition of this fact.

 Anders W's gear list:Anders W's gear list
Panasonic Lumix DMC-G1 Olympus OM-D E-M5 Olympus E-M1 Panasonic Lumix G Vario 14-45mm F3.5-5.6 ASPH OIS Panasonic Lumix G Vario 7-14mm F4 ASPH +21 more
Reply   Reply with quote   Complain
Post (hide subjects)Posted by
Keyboard shortcuts:
FForum PPrevious NNext WNext unread UUpvote SSubscribe RReply QQuote BBookmark post MMy threads
Color scheme? Blue / Yellow