Dpreviewmember: Fuji doesn't understand the market for bridge cams !
After shotting a lot with my Fuji HS20 and HS50 bridge cams, only in 8MP mode, (16MP EXR 1/2" sensor) I wouldn't expect anything but awful IQ for the S1's 16MP 1/2.3" even at 50% crops as is definitely confirmed by these samples. Unless you are an impressionist artist and enjoy mushy pictures ;-)
Why camera manufacturers don't realize that 16MP is a lot for a sensor of this size, 8MP would give better IQ specially in low light, faster processing of files and would be perfect for small to medium size prints as well as showing on 2MP fullHD and 4K TVs/monitors, which is what most buyers of this kind of cam actually do. So why insisting on MP counts ?
Depends how you look at it. I'd prefer my sensors to out-resolve the lens, and I prefer the natural blurring of diffraction to pixelation at large viewing sizes. It may just be a marketing ploy (wouldn't surprise me, really), but I don't think 16MP is ridiculous.
You need 8MP for a full 2MP of data. Studies have shown images downsampled from higher resolution sources can look better than images at a lower native resolution. The manufacturers know what they're doing.
And you can get better results from a high-res JPEG with more compression than a low-res one with less, as JPEG compression works better than simple pixel binning.
photofan1986: Images look good, for sure, but I see that as an amateur photographer, I really don't want a über-high megapixel camera.
@AlanG:I get that, but you can just as easily look closely at a smaller print, as well. It seems silly to me to fault a camera for being able to print larger. A lot of the talk here seems to imply (whether intended or not) that higher resolution is somehow a disadvantage.
Why should having more resolution ever detract more from an image? Missed focus or shakiness may limit the usefulness of the extra resolution, but it's not going to detract any more compared to a lower-resolution image viewed at the same size.
"...the higher the resolution, the lower the threshold of tolerance would be, imo."
That's an odd concept, in my opinion. Why do you feel that should be the case?
naththo: Significant amount of pixelation and aliasing in those samples. 35mm sensor are not meant to have much more than 20mp at least. Its way too much mp cramming into 35mm sensor is not good things. You need at least medium format to do 50mp or so which will show much clean image look. I compared to Phase One, Phase One came out much clean sample to compare with hardly much aliasing at all.
You must be comparing at 100%, which is the wrong way to go about things; you must compare at equal sizes. Set the image size to "print" in the studio compare tool for a better comparison.
What are you talking about? More megapixels reduces aliasing and pixelation. Aliasing is all about not having enough resolution.
Karroly: You forget to say that increasing the diameter of the "tubes" is like waiting longer under the rain : this will increase and make more equal the number of raindrops caught by each tube, thus reducing the "signal-to-noise" ratio. For sensors, it means, putting aside electronic noise, bigger pixels will catch more photons and average the random nature of light you are talking about here.
This is why, given the same resolution, a big sensor with big pixels will ALWAYS be better in (extremely ?) low light than a small sensor, even though there were no electronic (or thermal) noise at all...
However, in practice, the true question is, and your article does not give an answer, under which pixel size and scene brightness, does a pixel catch too few photons so that the difference between adjacent pixels becomes noticable ?Is this "noise" greater or smaller than the electronic or thermal noise ?In other terms, do we really have to care with the random nature of photons with today sensors ?
Okay, I see now that I've been misunderstanding what you've been saying!
What threw me of was when you said "No, you are wrong because you do not compare apple with apple. And the apple here is resolution. My hypothesis was : "given the SAME RESOLUTION, a big sensor with big pixels..." in reply to me talking about combining pixels of the sensor with more resolution. I assumed you were disagreeing with me, but it seems you were not?
I'm confused by what you're trying to say. But the fact of the matter is it's the difference in sensor size, not the difference in pixel size, that makes the difference.
Your insistence that comparing sensors with different resolutions is nonsense is absurd.
"So combining pixels with the small sensor will decrease the resolution compared to the big sensor... And comparing the noise level of two pictures that do not have the same resolution, in other words the same level of details is just nonsense."
That makes no sense. So if a new camera model has increased resolution, it's nonsense to compare it to the previous model? Do you always view/print images in proportion to their resolution? You never resize images for a particular output?
"If pixels where combined, that would negate the advantage of increased pixel count as it would decrease resolution."
True. But you cannot get more detail out of bigger pixels in good light, and you can perform smarter noise reduction algorithms than simple averaging.
"Also, each pixel is amplified individually in the sensor and it's this first amplification step that contributes the most to the sensors SNR."
Which is why I said "as long as the noise is scaled adequately." Every generation of technology has a sweet spot.
"You forget to say that increasing the diameter of the "tubes" is like waiting longer under the rain : this will increase and make more equal the number of raindrops caught by each tube, thus reducing the "signal-to-noise" ratio."
I think you mean increasing the signal-to-noise ratio?
"This is why, given the same resolution, a big sensor with big pixels will ALWAYS be better in (extremely ?) low light than a small sensor, even though there were no electronic (or thermal) noise at all..."
But there are more of the smaller pixels, so they can be combined, thus negating the big pixel advantage as long as the noise is scaled adequately. Smaller pixels just give you more detailed data to work with.
"Is this "noise" greater or smaller than the electronic or thermal noise ? In other terms, do we really have to care with the random nature of photons with today sensors ?"
In modern sensors, yes, in all but the lowest of light levels shot noise is the dominant factor.
junk1: Have manufacturers confirmed that signal levels (pixel size) are so low that randomness in the number of photons is noticeable?
I've always thought of larger sensors as capturing a larger signal, therefore hiding the electronic noise (and any other sources of noise), but how do we prove where the noise comes from (shot versus electronic).
Well, there's still some room for improvement in both areas. And the color filter makes for a pretty big performance hit, though there has not yet been a viable alternative. But even with a perfect sensor, shot noise is going to be a major factor at low light levels.
Pixel size does not really matter in regards to shot noise, the total amount of light does.
Manufacturers don't always give the details of their sensors, but there are people (many on this site) who do thorough testing to determine the electronic noise. In modern sensors, it's extremely low, and so shot noise is the dominant factor of noise in most circumstances.
ihv: Weird, CCD for aerials? Aren't shutter speeds more critical i.e. higher ISOs are more desired? CCD goes barely ISO400-800.
CCD does not mean a global shutter, nor does CMOS mean no global shutter.
xoio: "Canon is claiming it has a pixel count equivalent to the number of photoreceptors in a human eye."
A BS publicity comment... But it has already been widely calculated that for an image sensor to be equivalent to a human eye, it will have to be around 576 Mega pixels. Granted not all is used at the same time though.
What does "widely calculated" mean?And the human eye doesn't really work like a camera, so it's hard to compare; that 576MP number assumes equal detail from all portions of our vision, which isn't right. I've even seen estimates that the detailed portion of our vision is comparable to about 7MP.
WT21: I don't care at all about these, but it caught my eye that the ELPH160 and 170 are using CCD sensors?? What is that getting Canon? I'd almost think they were old sensors lying around, but at 20MP, they sound like new development. Of course, the limitation is immediately evident with movie mode of only 720 (maybe it's an avi file type?) Is there much upside to using CCD this day and age?
That caught my eye, too. 20MP CCD's doing 720p are at least two years old, so it wouldn't surprise me if Canon bought up some leftover sensors from some other manufacturer.
Albino_BlacMan: How long of an exposure do you need at 0.005 lux. Isn't that an imperative piece of information?
I assume some of the better sensors out there can capture some kind of image at that level if you leave the shutter open long enough.
f/1.4 at 30fps.