About image noise...
Let’s talk a bit about noise on pictures, which is one of the specifics of pictures made with digital cameras. We consider image noise every difference in the structure that is between the picture and the clear, ideal view of the theme (or we can also say that image noise is a random, usually unwanted, fluctuation of pixel values in an image). The most significant source of noise in digital cameras is the image sensor, while the difference visible on the picture is due to photon noise, which is given by the statistical uncertainty of the number of detected photons. In the photographical lighting range we can consider light streaming as continuous, but in fact it is not permanent, so no matter the camera measures X photons, that is never the exact number of photons that effectively hit the sensor, and a difference of a few dozens is enough to create image noise. Although noise increases as the number of photons do (the more light it is), but still, pictures taken in good lighting conditions are less noisy than those taken in darkness. The reason for this is that with lighting the number of photons increases faster than the noise itself, so the signal/noise ratio is enhanced, which makes noise become less evident – even though there is more than on a picture taken in darkness. A logical consequence of these is that noise level of cameras with a small sensor is higher, than in lens reflex cameras with a large sensor, since in a smaller sensor a smaller number of photons can be collected. And of course phones have a sensor smaller than even in entry-level cameras…
Picture taken with mobile phone in poor lighting conditions (SE K750i)
Night mode pictures are of lower quality also because of the so called dark noise. A part of the photons getting to the sensor hits an electron, which escapes from the attraction of the nucleus, thus creating a photoemissive effect. The photoelectrons are kept in the pixel by the control electrode’s electric field, and the quantity of the electrons collected this way is measured by the image sensor when it reads information. The problem is that pixels won’t contain only “useful” photoelectrons, but there will be ones created by thermal means, which are the source of dark noise. The number of these increases in direct proportion to exposure time and temperature, and we usually use a longer exposure time in poor lighting conditions, which is also a reason if image noise. By the way, this is quite a poor noise factor in cameras; it is visible only in the darkest areas.
Image noise at ISO200 (left) and ISO3200 (right) sensitivity (Pentax K100D Super)
Considering the topic of our article, it’s not these that are overly important, but noise due to sensitivity changes. As I have already mentioned, we don’t have to insert different kinds of films in order to change sensitivity, it’s enough to press a couple of buttons. There is another great war amongst manufacturers regarding ISO value, even the cheapest devices feature never before seen sensitivities (ISO 12800, etc.), but these are unusable without any exception. That’s because in case of digital cameras setting this value doesn’t alter the sensor’s sensitivity, but only amplifies the signal it produces, before digital-analog conversion. If, for example, we switch to ISO 800 from the base ISO 200 sensitivity, then the signal is amplified four times, and we get the same light level from quarter the signal, so from less electrons at the same time. The only problem is that the signal/noise ratio deteriorates and the amount of dark noise and readout noise increases, even though the latter one is already significant enough. In case of a small image sensor – like the ones in cell phones – the phenomenon is even more powerful, so when using a more advanced camera phone, that supports altering ISO settings, we can only alter the value in a narrow range, but I’m almost sure that the number of wars will soon reach this part too.
Noiseless, noisy, noise-filtered picture
Noise filtering is something also worth mentioning. Manufacturers try to enhance noisy pictures taken with small image sensors with different kinds of filtering algorithms, which usually aren’t very good. There are two wide-spread methods in cell phones: one of them is the one used in Sony Ericsson’s mid-range phones, which although eliminates every noise and artifact from the picture, but in turn it results in a blurred, oil-painting-like photo. Other companies are more discrete, they don’t eliminate image noise completely, just reduce it to an acceptable level, but this still results in smaller details getting lost. Due to some physical limitations of mobile phone they make noisy, filtered pictures, and unfortunately this is a fact, which is a partial answer to the main question of this article…
A cikk még nem ért véget, kérlek, lapozz!
