Hi Philippe,
image1(R) = 50% image1(R) + 50% image2(grey Ha)
image1(G) = image1(G)
image1(B) = 75% image1(B) + 25% image2(grey Ha)
Disable the
Use a single RGB/K expression option of PixelMath, and enter the following expressions for the R/K, G and B channels, respectively:
0.5*image1[0] + 0.5*image2[0]
image1[1]
0.75*image1[2] + 0.25*image2[0]
Leave empty both the A (alpha) expression and the Symbols slot. On the expressions above, note that the square brackets allow you to reference a particular channel of an image by its
index. Channel indices are numbered starting from 0, where the nominal channels are 0=red/gray, 1=green and 2=blue (when applicable).
You normally will have to disable the
Rescale result option of PixelMath when you're performing these combinations, but you can try out to see which option gives you the best results.
To apply this PixelMath instance, you have several ways. The simplest way is applying PixelMath to your original RGB image. If you do so, the resulting RGBHa combined image will replace your original RGB image.
Another way is creating a new RGB image with
Image > New. You should create a RGB color image of the appropriate dimensions. You normally will set the same dimensions of one of the RGB or Ha frames (the
Set As Active Image button is a handy shortcut), but note that PixelMath will interpolate source data automatically to match its destination, so you can specify any arbitrary dimensions. Once you have the new RGB image created, just apply the PixelMath instance to it.
A third way is telling PixelMath to create the resulting image. Open the
Destination section and enable the
Create new image option. Now apply the PixelMath instance to your RGB image, and you're done. Finally, you can also specify the width and height in pixels, select the RGB color space, and a sample format (usually 32-bit floating point). In this case you can apply the process globally (press F6, or click the blue sphere icon).
More if I have for example 4 images (Halpha, OIII, Hbeta, SII) and I want to mix some of them together to have a different RGB rendering than a single SHO lut I don't like.
Same as above, basically. In this case you have four images. For example, this is a 'natural color' (so to say) palette that you could try:
0.5*SII + 0.5*Halpha
0.15*Halpha + 0.85*OIII
0.5*Hbeta + 0.5*OIII
respectively for red, green and blue. Note that since the four frames are grayscale images, you actually don't need to specify channel indices, as 0 is always the default index for grayscale operand images.
Similarly, the expression at the beginning of this post can be replaced by this simpler form:
0.5*image1 + 0.5*image2
image1
0.75*image1 + 0.25*image2
because the default indices are being used with image1 for each RGB channel, and image2 is a grayscale image.
