View Single Post
Posts: 1,414 | Thanked: 7,547 times | Joined on Aug 2016 @ Estonia
#429
Originally Posted by mr_pingu View Post
Noise is solely based on sensorsize, amount of MP doesn't matter. More megapixel can* be even less noise when viewing on the same resolution / printsize. This is because more pixel can average noise out better.

Only way to improve imagequality is to gather more detail AND available light. Available light can be fixed with flashes. Xenon can deliver way more light than LEDs.

* The optics has to be able to deliver the amount of MP, if the optics doesn't there's no way then the noise can't be averaged out as much.
Let me jump into sensor size/MP discussion.

I don't know what is these days used in the phone cameras, but I expect it to be CMOS and a rather small one.

The cameras are having usually problems with lower light conditions, so let's focus on this case. Assuming that we have the same size of the sensor (CMOS or CCD chip), the larger size of a single pixel will lead to larger signal-to-noise ratio. This is due to a way the chips are transforming photons to digital signal. There are several parameters that determine the performance, but for know, we can concentrate on the noise sources. These include shot noise (photon nature of the light, not much to do about it), thermal noise (warm camera will get electrons from a thin air), and readout noise. Now, in low light conditions, you are frequently limited by readout noise. This is a noise describing physics of electron to digital readout conversion. In essence, when you feed to the readout circuit the same amount of electrons, you will be getting different numbers due to measurement error.

When we are limited by readout, signal to noise (SNR) for larger pixels vs pixels averaged in software is dramatically different. Namely, you will get considerably better SNR when you have a larger pixel than when you average after in software (including firmware...). See https://www.photometrics.com/resourc...s/pdfs/snr.pdf for simple explanation of physics behind and compare averaging of multiple frames vs single frame long exposure for readout limited condition.

For CCDs, you should be able to change pixel size on fly (assuming that it supports it) since you readout the whole chip through the same readout circuit. So, you could change your effective pixel size and adapt for lightning conditions. For CMOS, you are limited to software averaging since each column has a separate readout.

In the end of the day, at the lower light, I would always prefer larger pixels that lead to sufficient resolution. And, as stressed by @mr_pingu, to get the photons in with the decent optics.

As for flashlight, I would suggest to go for a faster solution - LED. You already have so many things to do and are able to differentiate yourself with hardware (keyboard) and software (making steps to deliver developer-friendly device with SFOS and Android), that I would suggest to focus on that this time and not on loosing time on a complicated flash.
 

The Following 7 Users Say Thank You to rinigus For This Useful Post: