Downgrading Megapixels

February 7, 2008 8:00 AM

Stupid-question-filter: Suppose you have a 10 MP camera and take your pictures at 3 MP. Would the quality be any different than if you were using a 3 MP camera at its top setting? (assuming the two cameras are otherwise identical)

Also: Do any image compression formats do a better job of reducing file size while preserving image quality than just taking a picture at lower resolution would?

posted by lullabyofbirdland to technology (10 comments total)

In answer to your "Also:" - JPEG-2000
posted by unixrat at 8:07 AM on February 7

No. The quality would be the same.

But your assumption isn't valid in practice. You're never going to find two cameras, 3mp and 10mp, that are identical except for the resolution of the final image. The image quality depends largely on the camera's optics, and on the sensor that digitizes light. See this.


Also: Just about any image compression format should do a better job of reducing file size while preserving image quality than just taking the picture at the lower resolution.
posted by qxntpqbbbqxl at 8:17 AM on February 7

lullabyofbirdland posted "assuming the two cameras are otherwise identical)"

Everything else being equal if the sensors are the same size the 3MP sensor will have larger pixels and therefor, in theory, produce a better image. However a lot is going to depend on the software being used to downsize the 10MP sensor.
posted by Mitheral at 8:23 AM on February 7

You should technically get slightly better results with the 10 MP because of the way camera sensors work. Any given pixel really only measures light intensity. Shades of gray. Too bad no one wants a black and white camera. So, to make it a color camera, they put a red, green or blue filter over each little pixel, in a specific pattern. Then the camera's circuitry turns it into full color. Each pixel's full color value is averaged out from looking at it's differently-colored neighbors. The process is called demosaicing. (FYI, RAW files are actually the data before this step.)

Demosaicing can cause artifacts in the image, especially when you're looking at tiny details that may only occupy a single pixel width. Your 10 megapixel image has these artifacts at full res, but converting it to 3 megapixel turns every block of 4 pixels into 1 pixel (3872 x 2592 vs 2048 x 1536). That averages out any artifacts and should give you a better image.
posted by smackfu at 8:24 AM on February 7 [1 favorite]

I would expect a noticeably better picture from the 10MP camera, for the reasons smackfu gives.
posted by teleskiving at 8:48 AM on February 7

I think a more appropriate question would be, "Would taking the pictures at 3MP be any different than taking the pictures at 10MP and then using software to resize them."
posted by qvtqht at 8:54 AM on February 7

However a lot is going to depend on the software being used to downsize the 10MP sensor.

True. Especially if you're using the in-camera downsizer, since they use simpler algorithms due to the relatively weak camera processor.
posted by smackfu at 9:29 AM on February 7

There is no answer to this. It's too vague. You'd need to site specific cameras & settings to have an answer. There are far too many variables involved to answer abstractly. If you want to see a recent real world example, you can check out the recent reviews of the Fuji Finepix f50fd (12MP) versus its nearly-legendary 6MP ancestor, the Fuji Finepix f30/f31fd here. In this case, scaling 12MP images from the 50 down to 6M to match the 30 gave results that were pretty much the same.
posted by chairface at 11:04 AM on February 7

Also: Do any image compression formats do a better job of reducing file size while preserving image quality than just taking a picture at lower resolution would?

Yes. A much better job. In my experience, for taking digital photographs and for scanning negatives, a high resolution image with jpeg artefacts generally contains far more detail than the same filesize achieved by being at a lower resolution without compression.
"Detail" meaning real information, such as "could you read the license plate of a car in the picture?"

There are some kinds of graphic art where I would trade the lower resolution for the absence of artefacts, but for the average photograph that is the wrong trade-off.

Furthermore (if you are not aware) there are lossless image compression formats, which reduce the filesize while preserving image quality completely. They work using algorithms much like those used in Zipping a file. Their limit is that they can't usually reduce the filesize to less than 50% its uncompressed size, whereas jpeg, which sacrifices some detail and introduces artefacts, can reduce the filesize a lot more.
posted by -harlequin- at 11:44 AM on February 7

Thanks for your help! I marked the most generally educational answers.
posted by lullabyofbirdland at 12:35 PM on February 8