@Alfredo beltran#57715 2x2 would mean 4x read noise
Hi Alfredo,
No, not 4x, since the 4 read noises are independent random variables. They do not add, it is their variances (power) that add. That is why binning in CMOS will still improve the SNR (assuming the original data is oversampled, and therefore the signal in the 4 pixels are correlated and not independent).
As for retaining color information, it depends on whether you asked for Mono-binning or "color" binning, when the processor in the CMOS camera is asked to bin the sensor data.
With color binning, you are not adding 4 adjacent pixels, but a binned red pixel is created from 4 adjacent red pixels in the Bayer array. The resolution drops drastically. It really only makes sense if the original image is very oversampled.
With mono binning, you are adding 4 adjacent pixels in a 2x2 array (like what you described), and therefore you loose the Bayer information, and as a consequence, also loose chroma information.
However, the resolution of the now-monochrome mono-binned image is not that much worse than the original Bayer image since the original color Bayer image has a reduced resolution compared to a monochrome camera anyway.
Chen
