What does 'bin' mean?

[DavidLeodis]

In the photographers technical photographic details with APOD images it often states such things as "1x1 bin" or "2x2 bin". I have tried to find out what bin means in this context but I have not been able to readily find out. I would guess that it is not just an astronomy photograph term so I have asked this question in the 'Discuss Anything' section. I would be grateful for any help. Thanks.

Edit added July 14 2011 at about 10:50 British Summer Time. 'Grateful' should have been 'greatful'. Oops, my mistake.

[orin stepanek]

I'm not hep on it either but it sounds like it has something to do with Astrophotography!
http://ciera.northwestern.edu/Outreach/ ... hotos.html

[rstevenson]

It's a general term referring to a way to quantize data. See this Wikipedia article.

Rob

[Beyond]

I just 'bin' there.Makes sense to me,but that's not really saying much.

[Chris Peterson]

In the photographers technical photographic details with APOD images it often states such things as "1x1 bin" or "2x2 bin". I have tried to find out what bin means in this context but I have not been able to readily find out. I would guess that it is not just an astronomy photograph term so I have asked this question in the 'Discuss Anything' section. I would be grateful for any help. Thanks.

A CCD or CMOS sensor has a native pixel count, such as 1000x1000. It is possible to bin the data so that multiple pixels on the sensor are combined into a single superpixel. If you use 2x2 binning, the data from the above sensor will be a 500x500 pixel image. There's really no such thing as 1x1 binning; that would more appropriately be called unbinned.

Why bin? Well, the most common case is where you are oversampled, meaning that the angle each pixel subtends on the sky is much smaller than the true resolution of the optics and seeing. By grouping adjacent pixels into superpixels, you better match the resolution of the system, and you also reduce noise. That's because an important noise source in dark areas is readout noise, which is injected into the data for each pixel read from the sensor. If you can combine the charge from multiple pixels before readout, you have less noise for a given area of sky. Binning can also increase dynamic range, since a superpixel holds more electrons (meaning it can collect more photons) without saturation.

For amateur astroimaging, you commonly see images made by combining a single unbinned luminance (B&W) channel with binned color channels. The binned color allows for lower noise at the expense of resolution, but our eyes don't see a loss of resolution in the color channels if the luminance channel is high resolution. This is the same effect that makes pictures in comic books look pretty good, despite the fact that the colors don't really register with the black ink very well.

[DavidLeodis]

Thanks to all who replied to my query. Your help is greatly appreciated. I now have an understanding of what 'bin' means. I noticed in one link brought up there was mention of 3x3 bin. If I understand correctly that means 9 pixels were reduced to 1. Presumably though there must be a limit to what amount of binning (? correct word) can be done before the image being processed loses too much detail.

[Chris Peterson]

Thanks to all who replied to my query. Your help is greatly appreciated. I now have an understanding of what 'bin' means. I noticed in one link brought up there was mention of 3x3 bin. If I understand correctly that means 9 pixels were reduced to 1. Presumably though there must be a limit to what amount of binning (? correct word) can be done before the image being processed loses too much detail.

Binning need not result in any loss of information. That's because if the native pixel scale is finer than the resolution imposed by seeing or optics, there is no more spatial information in group of four or nine pixels than in the individual pixels. But of course, as you note, if you keep binning larger and larger numbers of pixels, you inevitably reach the point where you will start losing spatial information.

I should also point out that the primary advantage of binning- the reduction of readout noise- only occurs if the binning is on-chip, meaning that the design of the sensor is such that the pixel charges can be combined before readout. This is in contrast to off-chip binning, which is functionally identical to resizing. To understand on-chip binning, you need to understand how a CCD works. It consists of an array of photosensitive electron storage wells (pixels), which are arranged in rows. As photons strike a pixel, it accumulates electrons. When the chip is read, those electrons are shifted along each row, bucket brigade style, with the charge from the end pixel of each row being shifted into a column of charge storage wells. That column is then shifted out to a readout cell and amplifier. So for a 1000x1000 element CCD, each row is simultaneously shifted once, and then the readout column is shifted 1000 times, and this is repeated 1000 times until all 1,000,000 pixels have been read (and cleared of charge in the process). When you bin 2x2, you shift each row twice, so each cell in the readout column contains two pixels worth of charge, and they you shift that column out two-at-a-time, so the readout cell contains a total of four pixels worth of charge. If you bin 3x3, you shift in groups of three, and the readout cell contains nine pixels worth of charge. This is what actually places the main limitation on binning size: while the number of pixels the readout column and readout cell can hold is usually greater than the number an individual pixel can hold, it isn't that much greater. Many sensors are designed so that their readout columns can hold twice as many electrons as a pixel, and their final readout cell can hold four times as many. What this means is that you can bin 2x2 without saturating something in the output path. But if you bin 3x3, you will definitely saturate if you have too many bright pixels in a superpixel. So usually, 3x3 and higher binning is only used for getting fast readout when aiming, focusing, and other cases where exposure and resolution are generally of secondary concern.

[DavidLeodis]

Thanks Chris. I would be wrong to say that I fully grasp your reply, but I do appreciate your help. David.