Please Explain Debayering

[Astrosurf]

I have an ASI120MC cam and prefer to capture in RGB with SharpCap. It defaults to RAW and I ended up capturing in that setting today. The vid is monochrome and covered in dots (Bayer?). I read in the manual that the image needs to be debayered, which Autostakkert can do. So, the CMOS chip has a Bayer pattern, so why debayer?

Is it better to capture in RAW or RGB? and what's the difference?

And as an aside, I can rarely get above 22fps using USB3 with my laptop. Any idea why it isn't faster? I know that increasing the Saturation slows the frame rate down which is why I try to keep it as low as possible.

Thank you!

Alexxx

[Helen]

Hi Alex,

Each pixel in a colour camera with a bayer matrix has a filter over it - to let through either red, green or blue.  They are grouped in four pixels and each 4 has 2 green, 1 red, 1 blue (to match the human eye's response to colour).  If you capture in raw, all the camera downloads and is saved initially, is a measure of the photons captured by each pixel.  It is quick as it just counts, it doesn't worry about what filter was over the pixel.  The debayering process takes the measurements and then assigns the colour to the measurement - this gives the colour output.  It is quicker to read /capture in RAW and then use software to debayer later.  You just need to know the colour pattern for each group of 4 - but trial an error is quick to sort that out if you don't know!

Helen

[Astrosurf]

Thank you!

[JamesF]

My preference is always to capture raw data because that's as close as possible to what comes off the camera sensor.  As Helen says, each row of data in such an image (usually) has filters in front of the photosites corresponding to each pixel, alternating green-red-green-red... and blue-green-blue-green for the next row.  Different manufacturers may arrange the colour filters (masks) in a different order, but ultimately they result in the same sort of thing.

Given the raw data there are many different ways to generate a full colour image.  Some of these are simple and fast but don't handle some image effects very well, whilst others create a more "realistic" image at the expense of increasing time for computation.  For example, some of the simpler algorithms don't handle sharp lines very well, creating a jagged edge, whilst more complex ones may do a better job.  If you have the raw data it's easier to go back and try something different at a later date.

As to why you can only get 22fps with your camera, there may be a number of reasons.  The first thing to check is that you say you have an ASI120MC.  Is that the USB2 version, or the later USB3 version?  My understanding is that the USB2 model may actually get a small performance boost when plugged in to a USB3 interface, but won't perform anywhere near as well as a true USB3 model.

James

[Astrosurf]

Thanks James. I'm not sure if it is USB3 to be honest. Will have to check.

But why is it called DEbayering?

[AngryDonkey]

23 minutes ago, Astrosurf said:

But why is it called DEbayering?

It is the process used to construct a full colour image form image data that has been created with a bayer matrix in front of the image sensor. So in a way it's reversing or removing (demosaicing) the pattern that is introduced by the filter matrix - hence the prefix 'de' (although in reality it's more a reconstruction process).

[JamesF]

Bayer was a chap who worked for Eastman Kodak who invented the array of red/green/blue colour filters (the Colour Filter Array, or CFA) placed in front of the photosites on the sensor to allow a colour image to be sampled in one exposure.  As Mike says, the process of turning the image created as a result of the Bayer filter back into a full colour image is therefore debayering, also known as demosiacing.

James

[Astrosurf]

10 hours ago, AngryDonkey said:

It is the process used to construct a full colour image form image data that has been created with a bayer matrix in front of the image sensor. So in a way it's reversing or removing (demosaicing) the pattern that is introduced by the filter matrix - hence the prefix 'de' (although in reality it's more a reconstruction process).

That makes so much sense! Thank you.