Pages: [1] 2 3   Go Down

Author Topic: How many pixels actually?  (Read 26973 times)

dalethorn

  • Guest
How many pixels actually?
« on: April 28, 2008, 12:12:49 am »

I use point & shoots, which I imagine few of you do. Since their pixel specs are usually way overrated, I wonder if there's a better way to rate them. For example, if I were to take a series of landscape pics under ideal lighting conditions, do minimal or no post processing, then make several copies of each at different sizes (say, 3072x2304, 2560x1920, 2048x1536 etc.), and view all copies at full size on screen, could I say that the True Effective Pixels is the smallest size at which I can plainly see all of the detail of the original, ignoring edge artifacts etc. added by the in-camera processing? I ask this because my highly-promoted 9mp camera appears not to have any real detail beyond about 3mp.
Logged

Tony Beach

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 452
    • http://imageevent.com/tonybeach/twelveimages
How many pixels actually?
« Reply #1 on: April 28, 2008, 02:49:30 am »

The real question is one of enlargement of the pixels, which are themselves limited by the lens and by your technique (whether you are using a tripod and mirror lock-up for instance).  Furthermore, the photosites that generate the pixels you get from your tiny little P&S sensor are much smaller than the photosites my DSLRs use to generate their pixels; add to that better lenses and more careful processing of those pixels (I always shoot in RAW for instance) and there is a dramatic difference in how much larger a print from a DSLR can be compared to a Digital P&S.
Logged

Dale_Cotton

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 588
    • http://daystarvisions.com
How many pixels actually?
« Reply #2 on: April 28, 2008, 06:52:38 am »

Dalethorn wrote:
Quote
I ask this because my highly-promoted 9mp camera appears not to have any real detail beyond about 3mp.
I suppose this varies from make to make and model to model, but in my experience resolution is the one area in which point & shoots have not been especially weak. I've used a variety of both point & shoots and dSLRs and have consistently found the amount of detail in the point & shoot images to be in the ballpark of their megapixel count.

This is confirmed by the careful testing in the reviews at dpreview.com, which gives resolution figures on the page just before the conclusion. For example, the 10 mp Canon G7 measures at 1775 horizontal LPH and 1850 vertical, while the 10 mp Canon D40 weighs in at 2100 and 1800, respectively. The only caveat here is that these are measured from the in-camera JPEGs, not from the RAW.

So if you're only seeing 3 mp of detail, you might want to consider whether you have a defective unit.
Logged

sojournerphoto

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 473
How many pixels actually?
« Reply #3 on: April 28, 2008, 07:06:02 am »

Due to the small pixel dimensions P&S sensors lose resolution due to diffrction earlier than dslrs (to date at any rate - pixel sizes are shrinking there as well). However, we have a 5Mp and a 9 or 10Mp P&S and both seem to be able to provide resolution that is consistent with the numbers. Against this is the noise level that means the files are less malleable, even at base iso, than an dslr file and they tend not to print as well at bigger sizes.

Mike
Logged

dalethorn

  • Guest
How many pixels actually?
« Reply #4 on: April 28, 2008, 08:36:01 am »

Correct me if I'm wrong (and my camera is not defective), but we seem to be missing the point. A lab can measure pixels and "resolution", but cannot see the difference between actual subject and filler, i.e. either "corrected" noise or interpolation. So back to the question - since the proof is in the viewing, am I wrong to suggest that reducing the image to the smallest size that retains the actual detail is proof of the real effective resolution?
Logged

Jonathan Wienke

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 5829
    • http://visual-vacations.com/
How many pixels actually?
« Reply #5 on: April 28, 2008, 09:30:05 am »

Quote
So back to the question - since the proof is in the viewing, am I wrong to suggest that reducing the image to the smallest size that retains the actual detail is proof of the real effective resolution?

No. The trick is to accurately define the threshold size below which detail starts disappearing, and account for resizing artifacts. Probably the best way to do this would be to shoot a resolution test chart, and see how much you can shrink it before the smallest resolved details disappear.
Logged

Plekto

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 551
How many pixels actually?
« Reply #6 on: April 28, 2008, 06:08:14 pm »

And it gets trickier, since those tests are almost always done with a black and white graph.  If you switch to color, the nasty Bayer filter rears its head.  Red and blue on a typical sensor are much lower resolution than the green channel.(typical distribution is 25%/25%/50% R/B/G)  As a result, it takes 7 or 8 of these sub-pixels to create a single full color point on the film.

The camera has to then use its internal software to interpolate and anti-alias what it's seeing(or in the case of RAW, the software has to do this)  Point and shoot cameras of course tend to have very poor methods of doing this and often won't shoot in RAW format. True, you can see some fine details around edges and such, but they tend to have off-color areas, moires, halos, and whatnot, because while the resolution is technically there, it's only partial color resolution.

So shooting color, you effectively have about 0.6-0.65x the listed resolution in actual full-color locations/true pixels in each dimension.  And, yes, their marketing departments lie - huge surprise there...  12MP is actually closer to 4.3-5.0 physical full color locations.  Which is why the Sigma at 4.6MP looks about identical to a 10-12MP CMOS sensor.  Yes, I know Sigma lists theirs as 14MP, but it's a true 4.6MP.

Try shooting a resolution chart printed in red, blue, and green.  I'm amazed that more places that review cameras don't do this.

http://www.ddisoftware.com/sd14-5d/
A review that points this out.  The problem isn't the lens or the camera, but that we need a new generation of sensors that are free of these problems.  I've nearly taken Foveon out of consideration, though, since they seem to be unable to get their act together and make a deal with Nikon or some other major player.

A true 8MP camera would be an ideal replacement for 35mm film.  Too bad Foveon and Sigma seem to be stuck in first gear in getting it to market.

EDIT: Why I said 8MP?
A typical D-Lab scans 35mm film at 3000*2000 fixed resolution.  This is roughly 2200 DPI.  400DPI Dye-sub as a result.  A more reasonable limit would be 2400DPI, though, which would net a 3400*2265 scanned area.  Almost all dedicated personal scanners default to this resolution as well, or very close to it.   This is roughly equal to about 420lines per inch on Dye-sub, which is where most experts agree that you don't gain anything - at least for color.  Most D-labs kludge it at 400 lines per inch and call it a day.  

So 2400DPI scanned is a good compromise and what we should aim for to consider 35mm film "dead"/rendered moot.

Anyways, 3400*2265 is ~7.7MP of actual true resolution.  That's 7.7 with a Foveon type sensor or about 5200*3500 for a Bayer pattern.(16MP at a 0.65 ratio)  It's not a lot different in terms of linear resolution - 5200 vs 3400(roughly like 300 DPI Dye sub vs 400DPI), but of course, the multiplier once you add in both dimensions makes it grow to silly numbers.  

That said, many DSLRs do give you 16+MP now and can give you film like results after they have been processed and tweaked. I said 8MP because it's not common to find sensors in exactly 35mm film aspect ratios these days, so you usually have to go to something like 3500*2400.

MF, btw, by the same standards, would require 30MP actual resolution, or closer to 60 with a digital back to replace 120 film.  The sensors are larger, and the software is generally better, making it closer to a 0.7 ratio instead of 0.65 or so.  I'd rate consumer level point and shoots at 0.6 or worse - which is why they look washed out and dull - there's just too little color data/saturation.  They just came out with 40MP digital backs, so 60MP or so and replacing 120 film entirely isn't that far off.

*note* - this isn't subjective measurements.  Most pros consider 120 film dead with 40MP, or close enough to where they don't care.  But from a technical standpoint, 60MP is about where you'd have to go to actually make it moot.  I give the industry 2-3 years to get to this point.
« Last Edit: April 28, 2008, 08:33:51 pm by Plekto »
Logged

dalethorn

  • Guest
How many pixels actually?
« Reply #7 on: April 28, 2008, 07:08:47 pm »

What got me into this was all or nearly all of the Internet camera site editors and pundits *insisting* to the mfr's that small cameras with big zooms should start out with wide angle (28 mm) so users can take better landscape photos. I insisted (to no avail) that cameras like the TZ5 would benefit more from a 35-350 mm range than 28-280, since landscapes would just show pixel smear at full view. The TZ1 in fact was 35-350. Unlike those pundits I suppose, I have the LLVJ 1 to 12 in hand in the 4-packs, and am well acquained with what a landscape can or should look like. Perhaps a few well-placed Internet postings reviewing these "landscape-ready" pocket cameras as they really are will help shake some of the camera mfg. spokespeople out of their lethargy.
Logged

Plekto

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 551
How many pixels actually?
« Reply #8 on: April 28, 2008, 08:38:58 pm »

Sounds good   I'd like to read it.

Oh - about Foveon...  I think the reason they aren't coming out with a better sensor is because Canon and the others beat them to it.(16-21mp).  If I was them, I'd be pouring money into a ~30MP digital back(they'd probably market it as 90mp - sigh).  That would  make a big impact on pros as well, since the technology seems more geared towards situations where color balance and saturation are important.  Most consumers just don't care by comparison.
Logged

sojournerphoto

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 473
How many pixels actually?
« Reply #9 on: April 29, 2008, 07:57:35 am »

I had a similar view of the need to have lots of pixels to equal even 35mm film, but when I bought my 5D I was astounded at just how capable it is in comparison with 35mm. Small sensor P&S cameras are very different and 10 or 12Mp on a small sensor isn't the same as 35mm film - but it's not just resolution that's lacking but the way the image is drawn - there's more noise (certainly at iso 100+ on mine) and much more depth of field.

Also, whilst it's certtainly the case that Bayer sensors do not resolve with the same efficiency as a sensor that could (efficiently) record all colours at each photosite, other people on here who are more knowledgeable than I have suggested that they are capable of resolving at about a 70 to 75% ratio. In part this may be because the world does not comprise single colours, but it supports the observation that a 4.6Mp foveon should be similar in resolution to an 8 or 9ish Mp Bayer camera. I've not tested just observing.

Motly, I would just say that there are some significant differences in the way small sensor cameras draw compared to 35mm film, and that it's worthtrying out a couple of dslrs as you might be surpirsed by the results.

Enjoy

Mike
Logged

dalethorn

  • Guest
How many pixels actually?
« Reply #10 on: April 29, 2008, 09:36:08 am »

Redundant reply text removed.
« Last Edit: April 29, 2008, 09:39:53 am by dalethorn »
Logged

dalethorn

  • Guest
How many pixels actually?
« Reply #11 on: April 29, 2008, 09:38:00 am »

Two thoughts here:

I never calibrated a monitor, but have had several in my vicinity for years - a few laptops and desktops with both CRT's and flatscreens. I've displayed my better images on the various screens and learned to make the adjustments that I felt would have the best longevity for proper color. True, the P&S images tend to have the washed-out look, but that's not the only problem. My question here is, do monitors and their video drivers have a problem analogous to audio gear, where "frequency response" isn't flat, but has spikes and resonances? Do any of the monitor calibrations correct for uneven color reproduction?

Second thought: When I bought my Nikon 8800 it had a 2/3 sensor, and a few months later I bought a Casio EXZ1000 pocket camera with a 1/1.8 sensor, which Casio upgraded to 1/1.7 with the EXZ1200. I took photos of the park service's welcome board at Bolsa Chica with both, side-by-side, and can swear that the Casio image was the equal of the Nikon in every way, given slight color adjustments. But today all of the "superzoom" 18x cameras are using smaller 1/2.33 or 1/2.5 sensors. This sounds like good news for DSLR users, of course. But is it really? Take a long look at CameraLabs.com to see who's buying most of the DSLR's. If this is the wave of the future, the future looks grim.
Logged

sojournerphoto

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 473
How many pixels actually?
« Reply #12 on: April 29, 2008, 11:12:31 am »

Quote
Two thoughts here:

I never calibrated a monitor, but have had several in my vicinity for years - a few laptops and desktops with both CRT's and flatscreens. I've displayed my better images on the various screens and learned to make the adjustments that I felt would have the best longevity for proper color. True, the P&S images tend to have the washed-out look, but that's not the only problem. My question here is, do monitors and their video drivers have a problem analogous to audio gear, where "frequency response" isn't flat, but has spikes and resonances? Do any of the monitor calibrations correct for uneven color reproduction?

Second thought: When I bought my Nikon 8800 it had a 2/3 sensor, and a few months later I bought a Casio EXZ1000 pocket camera with a 1/1.8 sensor, which Casio upgraded to 1/1.7 with the EXZ1200. I took photos of the park service's welcome board at Bolsa Chica with both, side-by-side, and can swear that the Casio image was the equal of the Nikon in every way, given slight color adjustments. But today all of the "superzoom" 18x cameras are using smaller 1/2.33 or 1/2.5 sensors. This sounds like good news for DSLR users, of course. But is it really? Take a long look at CameraLabs.com to see who's buying most of the DSLR's. If this is the wave of the future, the future looks grim.
[a href=\"index.php?act=findpost&pid=192474\"][{POST_SNAPBACK}][/a]


to your first question, yes monitors can be non-linear and if not calibrated give widely varying results. IMages that look fine on my monitor at home are horrible on my office lcd.

Can you expand on the grim future please?

Mike
Logged

dalethorn

  • Guest
How many pixels actually?
« Reply #13 on: April 29, 2008, 11:57:56 am »

Since mfr's are switching to smaller sensors on P&S's (note: 1/1.7 on pocket camera 2 years ago; 1/2.33 on large superzoom today), and since DSLR's are increasingly being marketed to non-serious photographers, and since marketing triumphs over reason and logic, I see a clear and unmistakeable trend to dumb down the new product releases.
Logged

sojournerphoto

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 473
How many pixels actually?
« Reply #14 on: April 29, 2008, 02:09:19 pm »

Quote
Since mfr's are switching to smaller sensors on P&S's (note: 1/1.7 on pocket camera 2 years ago; 1/2.33 on large superzoom today), and since DSLR's are increasingly being marketed to non-serious photographers, and since marketing triumphs over reason and logic, I see a clear and unmistakeable trend to dumb down the new product releases.
[a href=\"index.php?act=findpost&pid=192502\"][{POST_SNAPBACK}][/a]


I saw a couple of people in a shop the other day and one was saying to the other, 'I really should get of of those dslr things'. Clearly just the next must have after the P&S, and probably less suitable for their needs.

Dumbing down is the inevitable result of turning items into commodities. On the other hand there remain lots of people who are willing to pay for high quality equipment. I hope though that it doesn't turn into the same esoteric melee as high end audio, where magic seems to be half the equation:)

Perhaps strangely, I have something of a problem with the UKs foody trend, where supermarkets sell 'special' food at higher prices. Usually the special food is what they should really have been selling all along, unlike the cut price dumbed down food that the british consumer expects to pay next to nothing for.

\rant mode off

Mike
Logged

Plekto

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 551
How many pixels actually?
« Reply #15 on: April 29, 2008, 08:32:12 pm »

Quote
I had a similar view of the need to have lots of pixels to equal even 35mm film, but when I bought my 5D I was astounded at just how capable it is in comparison with 35mm. Small sensor P&S cameras are very different and 10 or 12Mp on a small sensor isn't the same as 35mm film - but it's not just resolution that's lacking but the way the image is drawn - there's more noise (certainly at iso 100+ on mine) and much more depth of field.
True.  The software and in-camera processing is getting to be better, but it's really a kludge.  In a perfect setup, none of that would be required.  Thankfully, though, most good DSLRs are all at the 16MP or so range and have nice, big sensors as well.  They really do look a lot like film.(saturation and dynamic range and so on aside)  That the 5d can almost kludge its way to look like film is impressive.

Quote
Also, whilst it's certainly the case that Bayer sensors do not resolve with the same efficiency as a sensor that could (efficiently) record all colors at each photo site, other people on here who are more knowledgeable than I have suggested that they are capable of resolving at about a 70 to 75% ratio.

With the move towards smaller and smaller sensors, I see the exact opposite happening.  Some pocket AF cameras I'd rate as closer to 50% due to tiny lenses and microscopic sensors.  70% is about it for a Bayer type sensor without software and filters and such getting involved in the mix.  It's a technological problem inherent in the design itself.  Much like how you can't ever get 50% efficiency out of a 4 stroke engine and never will.  To get higher output, you need a turbine or other technology.  And the same holds true for cameras.  The fact is that the Bayer technology has pretty much run its course and we need a major new method.  One that captures the dynamic range and saturation of film more accurately, since resolution is at or close to being solved.

Foveon looks stunning, but the resolution is closer to 110 film.  I'm no longer expecting anything meaningful from them, though.   Fuji has a dual layer sensor, though, which appears to fix the dynamic range problem.  This looks like it might be a partial solution.  The photos it takes are noticeably more realistic looking than most cameras, since it is basically bracketing - but doing it in-camera off of a single shot.
(two sensor grids overlapped - each set a stop or two apart from each other, then blended, since there's 0 pixels out of place/it's taken at the same time)

http://www.outbackphoto.com/reviews/equipm..._S5_review.html
Looks promising, but it's only ~6MP.  We need 12MP or more.  But it does seem to be better than the Sigma by a tiny bit, mostly because of the extra resolution.   I'd rate this as equal to about 70% on the above scale.  It has a better saturation and dynamic range - more like a typical DB.  But it needs more pixels as well to really replace film and also compete with the "big boys".

P.S. look at the second set of photos.  There's no way a typical DSLR would see anything inside that upper floor behind that glass.  And, it uses Nikon lenses.  I hope they make a 12-16MP version of this soon.

http://www.bythom.com/s5review.htm
Another review.

**
"Most interesting is that the expanded dynamic range all happens above middle gray, which is exactly where most DSLRs have their least capability (I've written before that most DSLRs are asymmetrical in their ability to capture values below and above middle gray). If you're dealing with wedding dresses, white bird plumage, or anything else that has detail at the top of the highlight range, the S5 Pro has the ability to pull in two more stops of that."
**

They have a new camera out this summer.  We'll see if they fix some of the problems .
« Last Edit: April 29, 2008, 08:38:55 pm by Plekto »
Logged

Tony Beach

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 452
    • http://imageevent.com/tonybeach/twelveimages
How many pixels actually?
« Reply #16 on: April 29, 2008, 10:34:41 pm »

Quote
True.  The software and in-camera processing is getting to be better, but it's really a kludge.  In a perfect setup, none of that would be required.  Thankfully, though, most good DSLRs are all at the 16MP or so range and have nice, big sensors as well .
[a href=\"index.php?act=findpost&pid=192571\"][{POST_SNAPBACK}][/a]

I disagree with practically every assertion you have made in this thread.  My D300 has a 12 MP DX sized sensor and I would stack it up against 135 format film without any worries.  The resolution is excellent and the DR is a comfortable 8.5 stops.  The biggest problem with Foveon sensors is that they get noisy at relatively modest ISOs; which is another way that my D300 kicks butt on both Sigma DSLRs and film.
Logged

sojournerphoto

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 473
How many pixels actually?
« Reply #17 on: April 30, 2008, 04:21:09 am »

Quote
Thankfully, though, most good DSLRs are all at the 16MP or so range and have nice, big sensors as well.  They really do look a lot like film.(saturation and dynamic range and so on aside)  That the 5d can almost kludge its way to look like film is impressive.

[a href=\"index.php?act=findpost&pid=192571\"][{POST_SNAPBACK}][/a]

That would be just 2 dslrs with 16+Mp and full frame sensors at present - and one is out of production.

In my experience the 5D is comprehensively better than 35mm film for actual prints up to about 30 by 20. At larger sizes the differences become more apparent, but for most work film doesn't win out.

Mike
Logged

dalethorn

  • Guest
How many pixels actually?
« Reply #18 on: April 30, 2008, 07:51:58 am »

The main point of this thread was to discuss the mfr's design trends, and explore the outer limits of sensor capability. Certainly DSLR sensors won't capture anything comparable to scanning backs, so they have a long way to go. My interest is in smaller cameras, and my opinion is that given the *negative* trends there, the implications for DSLR's are also bad, due to market pressures and the dumbing down of the equipment in general. You may see progress for awhile, but things can change. You can get a "superzoom" mini-camera for handheld wildlife shots at ~500mm that will produce a usable image if the lighting is good, so try to duplicate that with a DSLR - handheld. You might, but it's going to be awfully big and heavy. And I don't see any relief there. It did happen in music players, it didn't happen in laptop computers, and I bet it won't happen in DSLR's.
Logged

Tony Beach

  • Sr. Member
  • ****
  • Offline Offline
  • Posts: 452
    • http://imageevent.com/tonybeach/twelveimages
How many pixels actually?
« Reply #19 on: April 30, 2008, 10:53:40 am »

Quote
Certainly DSLR sensors won't capture anything comparable to scanning backs, so they have a long way to go.

You can get a "superzoom" mini-camera for handheld wildlife shots at ~500mm that will produce a usable image if the lighting is good, so try to duplicate that with a DSLR - handheld. You might, but it's going to be awfully big and heavy. And I don't see any relief there.[a href=\"index.php?act=findpost&pid=192652\"][{POST_SNAPBACK}][/a]

Your expectations regarding DSLRs are unrealistic.  Frankly, even a DX or APS-C DSLR with a 400mm VR or IS lens attached to it will spank any P&S with a "superzoom" lens.  The extra weight of a bigger sensor and bigger glass you get using a larger format is unavoidable if you want to increase image quality and to gain more DOF isolation.
Logged
Pages: [1] 2 3   Go Up