Nokia Unveils Symbian 41 Megapixel Camera Phone
blogs.wsj.com
blogs.wsj.com
To be serious, I'm more of a physics guy. Photon "shot" (think shotgun with fewer and fewer pellets) noise determines already the lower limit of noise and (having a certain fixed shutter speed which is defined by avoiding movement) the only way you get more photons from the same subject is more lens area.
Maybe you can filter a little better if you have more pixels. I'd imagine the dynamic range sucks though if each pixel is so small that it saturates quickly.
So, if you're shopping for a camera, what determines how your photos come out in anything but the brightest sunlight is actually the lens. You should be shopping primarily for a good (read large aperture) lens, with a decent camera attached / bundled. Or if you want to go further, a system that is offering good lenses.
Well, if you replace a single sensor with four sensors that saturate more easily, observe that the chance that all four sensors saturate may be less than the chance that the original, larger sensor saturates. Answering the question, "how much light actually fell on these four sensors" becomes a more difficult problem and may yield greater error but is still tractable.
For you with a math background:
Let X_i be the amount of light that actually falls on sensor i, for i=1..4, and let e_i be the amount of uncorrelated noise. The sensor readouts are Y_i, where Y_i = min(X_i + e_i, sat) and sat is the saturation. Let X=sum(X_i). We know that X and each X_i follow approximately a Poisson distribution and we wish to know the parameter of the distribution for X.
Obviously, if Y_i = sat for all i then we are SOL. However, if not, then we can always use some kind of reasonable estimator for the desired parameter.
Your camera software chooses the shutter speed, aperture and iso sensitivity according to some formula like average lighting in the scene. But the real world has such huge dynamics so that basically either the whole sky is washed out or then you can't make out the person's face.
So basically all the pixels where the sky light falls on, oversaturate. If you lower the shutter speed, sensitivity and aperture so that the sky stays just below saturation, then you conversely catch very very few photons from the person. The person may appear pitch black or at best, really noisy, since the photons are so few. With noise, you often actually get color noise since because of the Bayer mask, the single photons hit random color pixels here and there. It looks really bad and cheap. Actually, that's the tell tale sign of low dynamic range and tiny pixels.
What do people do with even 12 megapixel 4000x3000 images? I'd settle with 2000x1500 - still greater than any resolution on any of the screens I own, and the dynamic range could potentially be better. The full res images look really bad and noisy anyway.
Actually, most of internet's picture content is just too large photos which for the most part don't contain any useful information: below some resolution you are just recording noise. I've set my compact to 5 megapixels.
Yet we haven't seen the end of wasted bit proportion. Phones recording full HD video and later 4K and 8K will explode the data demands a thousand times. Everybody will watch those resized to much smaller resolutions too.
> So, if you're shopping for a camera, what determines how your photos come out in anything but the brightest sunlight is actually the lens.
Actually I think that sensor is especially important in low light situations, because you need high sensitivity and low noise. So imho when the lens is good enough then what determines how your photos come out in anything but the brightest sunlight is actually the *sensor.
The angular resolution can be computed from the actual size of the aperture in mm (and the wavelength of the light you are interested in) and you do not need to know anything else. So my RZ67 has a maximum angular resolution of about 1 arc second, the D40 is limited to 4 arc seconds, and the iPhone is limited to 32 arc seconds.
The lenses are roughly equivalent in FOV, although the iPhone is slightly wider so it's actually a bit better than the numbers above indicate. The high resolution of the better cameras is also more limited by shake, and the actual aperture used will often be smaller; whereas an iPhone will be used more often wide open. Small lenses are more difficult to manufacture well, and lenses used wide open reveal a breadth of lens defects that a more reasonable aperture does not.
But f/2.8 is still unimpressive if f is small.
Let's assume that the aforementioned iPhone takes 8MP 4:3 images, ie 3266x2449 size. (I'm too lazy to check it's actual specs). To have similar angular resolution per pixel -ratio, the Nokia 808 needs to have 2.4 times higher width and height, right? Ie 7838x5877, which is 44 megapixels.
Based on this I conclude that that having 41 MP sensor on the Nokia is more sensible than having 8 MP sensor on iPhone. But this kind of analysis is probably horribly flawed, so instead I conclude that calculating theoretical angular resolutions is meaningless.
I first thought the sensor size must be a mistake, 1/1.2" huge, it is larger than any point&shoot camera (biggest ones are 1/1.5" like Canon G1-X) and its area is ~60% of normal DSLR sensor area. But still the 41MP sounds huge, it remains to be seen how much that MP count affects the actual image resolution. Maybe Nokia should just start making P&S cameras if it's really that good.
By my standard, it's 90% the quality of a u4/3. 81% the area. Close to a full frame D5 than an iPhone.
It is quite bulky.
What the hell, sensationalist title?
Full-res samples: http://cdn.conversations.nokia.com.s3.amazonaws.com/wp-conte...
It really sounds pretty neat.