Help needed please with GREEN !!!!!!

[Trevor N]

I recently purchased an Altair 138C camera. Seems to work very well and impressed with the detail in the images. I cannot however seem to stack these without the results being seriously GREEN !. I've saved using altair capture as fits and used dss to stack. I've opened the "RAW FITS" dialogue and changed to RGGB but to no avail. I guess I'm missing something here but after many attempts I'm going no where. Any help greatly appreciated. Trevor

[vlaiv]

First check if you got your debayer settings right.

Most of the time, LP is dominant in red part of spectrum and most people have issues with red cast rather than green. I'm assuming you in fact have Altair 183C rather than 138C and above is typo - altair website says proper bayer pattern is:

GBRG

It is either that or GRBG edit: RGGB (image can only be flipped upside down by software but not left to right, and that depends if one observes Y coordinate to go up or Y to go down. X always goes to the right).

If you are sure about your bayer settings and get those confirmed, here is "in depth" explanation on how color works and how to get proper color balance. Even when you get your debayer settings right, it is important to understand this.

Each sensor has different sensitivity curve regardless if it is mono + filters or OSC sensor. R, G and B from sensor do not map to R, G and B for display directly. Here is what published sensor response looks like for 183 model:

And here is what "response" for sRGB (color space most likely to be used by computer and therefore you should transform to this color space when working with computers / computer screens):

Now, don't be confused by the fact that there is "negative" part of the curve in sRGB matching functions. sRGB color space has lower gamut than human vision and three primaries used for sRGB (type of blue, type of red and type of green) don't allow for all colors to be reproduced. Here is sRGB color gamut compared to human vision:

sRGB color space can only display section in inner triangle, while humans can perceive everything that has been shown as color - much larger "area" in this chromaticity diagram.

You need to "subtract" some light to get things that eye can see but sRGB can't reproduce. In any case, that should not worry you much.

Important thing to note is that red and blue are "higher" then green in this graph. If you examine 183 spectral response - green is higher than other two. Raw color information from the sensor is not going to represent true color. You need to do color calibration to transform raw color information into sRGB color space. Most people don't do this, and they adjust color balance as they please (reduce green, boost blue and things like that - until they get result that they want). There are other ways to accurately transform raw color into sRGB (linear - it needs gamma adjustment later) color.

One of the ways is to do star color calibration. PI has script for this for example. Another way is to use precomputed matrix (from someone who did general color calibration for this sensor - this works only if you use same filters as that person), or alternatively doing color cards. You can purchase standard color cards that you can use for color calibration. In fact that is what they are designed for. Photographers use them to calibrate equipment. It looks something like this:

you take such card, provide uniform daylight illumination to it (or specific illuminant like D65) and shoot a picture of it with your sensor. From actual raw color values and expected color values for such chart you can derive transform matrix.

Btw, transformation is as simple as channel mixer and it goes something like this:

sRGB_red = a * raw_red + b * raw_green + c * raw_blue

sRGB_green = d * raw_red + e * raw_green + f * raw_blue

....

so you need nine numbers a, b, c, ..., i that represent your color transform matrix.

Bottom line is - don't expect raw color information to be properly white balanced, you need to do white balance on your data to get good results. Some ways of doing color balancing will produce accurate color results (star color calibration, color cards, ....), while others will be only approximate.

 

Edited October 24, 2019 by vlaiv
Wrong bayer pattern quoted for flipped image

[Trevor N]

Thanks Vlaiv. That should help a lot. Much appreciated. Trevor