2

I have pictures from three sources, a Galaxy note 3 phablet, a Ricoh WG-M1 sports cam, and a Canon EOS 600D.

I did set the EOS 600D to store as jpeg, and my pics are typically between 2.5 and 4.5 MB, and size is 3456x2304.

It seems the WG-M1 brings me pictures with less variation, all being around 4.5 MB, size is 4320x3240.

The Galaxy Note 3 brings me pictures between 3 and 3,5 MB and size is 4128x3096.

Obviously the pictures from EOS 600D are much better looking, on my screen at least.

So I wonder, could it be the case that the two other --lower quality-- devices are just saving too many pixels? Or maybe I should check again the setting on the EOS 600D and use a higher resolution? (But then, how come do these pics look better (on screen) than those of higher resolutions?)

  • If it's on your screen, 800x600px is plenty. That's less than half a megapixel. It's not about the megapixels. – user4894 Feb 16 '16 at 22:56
  • These pictures are not intended for screen only, or web only. Some are family portraits and events I'd like to be proud of in a few years, and print for family use, some are personal itches like pictures of moisture on walls or reflections in buses windows. I even sold one already! – gb. Feb 17 '16 at 3:03
6

To a very large extent, pixels don't matter - or at the very least, more pixels won't make things significantly worse.

What does matter is the size of the sensor in each camera - the EOS 600D has an APS-C sized sensor with a total area of around 330 mm^2, while the Note 3 has a 1/3.2" sensor with a total area of around 16 mm^2 - or in other words, about 5% of the size. That means that the EOS 600D can gather about 20 times as much light (depending a bit on the lens you're using) and so needs to amplify the signal a lot less in order to get a well-exposed photograph. Amplification means noise, and that's why the Note 3 looks bad. The Ricoh has the same sized sensor as the Note, so things will be very similar there.

Smartphone manufacturers will often attempt to compensate for the noise from the small sensors by very aggressively applying noise reduction algorithms to the images. However, these aren't magic and often mean you end up with "smeared" pictures instead of noisy pictures. With a dedicated camera, you often have some sort of control over the amount of noise reduction applied so you can choose, but for many smartphones it's fixed.

  • CMOS size is important only for sensitivity. Lower light - less noise. However, there are other factors as well. – TFuto Feb 16 '16 at 15:08
  • Err, I didn't mean that more pixels result in worse quality. My question is: Handling smaller files is better for other reasons (backup size, loading speed), so maybe I can reduce the resolution or increase compression for pictures taken by sports cam or smart phones. – gb. Feb 17 '16 at 3:00
  • @gb That was my interpretation too, since it's a fairly common belief. (See this question.) You might want to edit your question to make this more clear. – mattdm Feb 17 '16 at 12:48
  • @mattdm not sure how to edit, if you have an idea please suggest. – gb. Feb 19 '16 at 2:41
  • I asked another question photo.stackexchange.com/questions/74900/… – gb. Feb 19 '16 at 2:53
1

Your effective resolution is affected by:

  • your expected SNR from the sensor (signal-to-noise ratio, how much noise you accept),
  • sensor non-linear response, dynamic range, color reproducability,
  • sensor resolution,
  • lens resolution and contrast
  • expected loss factor from the file format.

Using the same quality JPEG setting:

  • SNR is determined by the sensor size, the CMOS technology, amplifier + ADC quality,
  • sensor resolution is determined by pixels by row/column,
  • sensor non-linearity is usually kind of given, sometimes compensated against, sometimes adjustable,
  • sensor dynamic range, color reproduction ability is usualy given, but changes with ISO, also with white balance in JPEG,
  • lens resolution and contrast is determined by objective design and build (see this link for more information).

Re: "So I wonder, could it be the case that the two other --lower quality-- devices are just saving too many pixels?"

It may very well be. If your sensor is small, the dynamic range is poor, the lens resolution is worse, the noise is higher - you are saving just too many pixels.

This is the same problem with a 40 MP camera using a lens designed for a 8 MP sensor. (Lens resolution is too low). Or using a 40 MP camera with a plastic toy lens (lens quality is poor). Or saving pictures into 24-bit RGB formats when the sensor can only digitize 5 bits per each color channel (dynamic range/ADC resolution is worse than file format).

1

Here's a more practically-oriented take on the problem:

  1. From the camera you want to check, find a photo of which you know that it's as sharp and in focus as possible with this device.
  2. Create a couple of scaled-down versions, e.g. to 50%, 60%, 70%, 80%, 90%.
  3. Scale up each scaled-down photo to the original size.
  4. Compare the re-upscaled photos side-by-side with the original photo in some image viewer with according functionality (or just open them and alt-Tab back and forth)

After this comparison, you should be able to tell from which scale-down factor upward virtually no resolved data is lost, as the according re-upscaled image will not show less detail than the original photograph.

To give only a very broad idea, I did this with a 16MP DSLR and got something like 75% (meaning width * 0.75 and height * 0.75, not MP * 0.75).

Note that this obviously depends that the camera really achieved its best in the original sample photo (in the focused regions), and there are many ways of not getting it perfectly right (aperture too wide or too narrow, too high compression, soft lens, next to unnoticeable shake,...).

  • This is only true if downscaling then upscaling back to 100% would not decrease perceived sharpness. However, this is not true, there is always loss of detail during downscaling. E.g. create an image of a fine pixel pattern, then add random Gaussian noise, and call the composite "X". Then downsize, then upsize it back to 100%, call it "Y". Substract "X" from "Y", and apply auto brightness+contrast. You will see all the mismatches between the two images, created by only the upsize/downsize algorithm. – TFuto Feb 17 '16 at 16:30
  • In general, any downsizing is a sort of low-pass spatial filter, removing fine details. Depending on how smart the upsizing algorithm is, it can guess missing fine details, but usually it creates artifacts. – TFuto Feb 17 '16 at 16:32
  • I guess you're right about that, but in photography I feel (YMMV) resolution is more about what objects and patterns a human can see or not see in a real photograph, than a hard definition of pixel resolution of a digital image. Even if the re-upscaled picture is a little soft, it will still be possible to determine if human-perceptible details were lost. – S.V.P. Feb 17 '16 at 19:40
  • You are right, but if you have high resolution, it is very hard to determine the loss of sharpness/loss of detail at 100% zoom. You would have to print the image at actual resolution in poster size or scan through large areas at 100% on monitor to have an idea how much detail is lost, and it is not easy to quantify, either. – TFuto Feb 18 '16 at 18:45
  • Good idea, is there an automated way to do that? See my new question here photo.stackexchange.com/questions/74900/… – gb. Feb 19 '16 at 2:58
1

I think Sébastien's answer is as good as it gets. But then again to really know if your camera has captured the highest resolution possible, you should use it on a sturdy tripod (or something where you can depend on that camera-shake is completely outnumbered). Do not rely on image stabilizers here. In fact: switch them off if possible! (Use it for all cameras that you compare in the test; thus gaining continuity).

I prefer to use natural textures for these kind of 'tests': grass, mountain-walls, trees, leaf, etc... Nature has an infinite amount of details! No test-card can compare to that.

As in resolving power, do not underestimate the importance of the right exposure: underexposure will lead to loss of detail (and obviously overexposure!).

Also, be careful with compression, because sometimes details get lost where you do not expect at first glance, mostly (colour-)textures (eyes, skin, fur, leaf, wood). When they disappear, the photo gets an unnatural appearance. I would like to point out to Chris that some cameras do let you choose the compression factor (as in low, mid and high compression), next to the resolution (of the JPEGs that is).

Keep in mind that while taking the photo and while post-processing details can get lost.

Finally, I choose the resolution - in post, I always shoot RAW - that suits the result (1280 pixels max. for web, full resolution for print), regardless of the real resolution as a result of lens quality for instance.

But perhaps that is a way of thinking, because in the past I was unable to change the film size afterwards...

  • Welcome to photography.SE Jos :) I don't think you are answering the right question, which is How to know if a picture has too many pixels?. Your answer seems to explain how to take a correct image of a highly textured subject without losing details. – Olivier Sep 18 '16 at 10:08
0

The level of compression in a jpeg is adjustable in software (including the software running in the camera, that you have no control over). This is in addition to the pixel count, and for the range of sensors you quote, will dominate any effects due to the number of pixels (you could try this out in your image editor of choice).

It is possible to save jpegs targetting a particular file size. This is handy for some web applications, where the photography is secondary to the main prupose of the website. This isn't as common as it used to be but some mobile operators recompress with a file-size target.

What I think is more likely is that the image-processing software on the cameras (including the rather aggressive noise-reduction @PhilipKendall describes) tends to produce similarly-sized jpegs. A very high level of random noise could have the same effect of making file sizes consistent, but I suspect the noise would be unacceptable long before this. You could have a play with comparing photos of white paper, black paper, and something complex (a printed photo for example) under consistent lighting conditions. You'll almost certainly see variations in noise and file size for any given camera. The auto exposure will be a confounding factor that you probably can't avoid except on the EOS (by using manual mode).

Your Answer

By clicking "Post Your Answer", you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.