Page translated by Claude — switch to Italian to read the original article.

INTRODUCTION

In recent years, astronomical photography with “aesthetic” purposes has spread widely among enthusiasts thanks to the advent of increasingly sensitive electronic sensors at affordable prices.

The use of RGB filters has made it possible to obtain spectacular color images with black-and-white cameras; the downside is that these filters often “steal” a lot of light and therefore, in order to obtain good signal-to-noise ratios, very long exposures in the three colors are required.

The alternative is to add a fourth filter to the arsenal: a completely transparent filter (Luminance) that captures the integral light of the photographed object.

In this way it is possible to take long exposures in L, which will contain the detail, and decidedly shorter exposures in RGB that will be used solely for the chromatic information:

This technique is called LRGB composition; as we will see below, this type of composition requires that the images to be merged be non-linear, and this, in some cases, constitutes a limitation.

This tutorial describes a technique for merging the L image with RGB in the linear stage and assumes that the reader has a clear understanding of how standard LRGB composition works.

As astrophotographers well know, an image fresh out of the electronic sensor has very little to show: the image is linear, which means that the signal on each pixel is linearly proportional to the brightness of the source that illuminated it.

Astronomical objects, such as galaxies and nebulae, are usually extremely faint compared to the field stars, so the monitor is not able to show the faintest, but most interesting, part of the image, which will appear as a substantially black field with a few sporadic white dots corresponding to the brightest stars.

To make an image visible it is therefore necessary to apply a non-linear histogram transformation, acting on the midtones and giving greater emphasis to the faint part rather than the bright one, an operation called delinearization.

The great advantage of PixInsight over other software used by astrophotographers for image processing is that, thanks to a particular process called ScreenTransferFunction (in short STF), it is possible to work with linear images without delinearizing them beforehand; in practice ScreenTransferFunction applies a non-linear histogram transformation only on screen, leaving the linearity of the image unaltered.

For this reason, those who use PixInsight prefer to keep images in linear form as long as possible.

Unfortunately, LRGB composition requires that the color image (RGB) and the Lightness image be delinearized before composition, because, in order to perform the merge, the CIE Lab or CIE Lch color space is used, which are non-linear spaces (the function that transforms the R, G, B values into the corresponding L is not linear) and therefore the use of linear images generates colored artifacts on the final image.

That is why I tried to develop a procedure to perform the merge in the linear stage.

The method is based on the fact that the HSI color space, unlike CIE Lab and CIE Lch, uses a linear function to calculate the Intensity I from RGB: in practice I=(R+G+B)/3

We can call the composition obtained in this way an IRGB image

THE METHOD

As with the standard LRGB technique, we start from the four base images: the three color channels, R - Red, G - Green and B - Blue and the luminance channel L: all the images being linear and aligned with each other.

In case there are brightness gradients due to light pollution, they can be removed individually on the monochrome images, or all at once after the RGB composition.

The first step to perform is the creation of the linear RGB image:

• open the ChannelCombination process.
• select the RGB color space.
• load the three frames into their respective positions.
• press the button with the blue circle.

In this way the linear color image will be created, on which it may be necessary to perform gradient reduction (if it has not already been done on the individual R, G, B components) and, above all, color calibration.

The RGB image might be very noisy, since - often - fewer frames are captured compared to the luminance, so it is necessary to proceed with noise reduction in the linear stage.

There are several tools in PixInsight to accomplish this task: among the most efficient on linear images are TGVDenoise, MultiscaleMedianTransform, MultiscaleLinearTransform and the newest one, the script MUREDenoise (to be used as the very first operation on the black-and-white images of the individual channels).

If it is not clear how to perform color calibration and gradient reduction, I invite you to read the initial part of my article Example of Processing with PixInsight.

Since most of the visible detail will be found in the luminance channel, the noise reduction on the RGB can also be very aggressive (without however completely erasing small details such as the faintest stars).

At the end of the procedure we will have an RGB image without gradients, perfectly calibrated, with little noise (although slightly poorly defined)

The starting RGB image after performing gradient reduction, color calibration and noise reduction. The image is still in the linear stage with an STF applied

At this point it is necessary to prepare the L image; in this case too it may be necessary to reduce light-pollution gradients, perform a moderate noise reduction or other operations to accentuate the details such as deconvolution.

Once this is done, the image must be prepared for merging by matching the brightness levels of the L image with those of the RGB image; in fact the luminance is often captured with different exposure times or binning compared to RGB, and moreover the L filter, letting through the entire visible spectrum, is much more sensitive to light pollution than the RGB filters.

The difference between LRGB composition and IRGB composition lies precisely in the linearity of the image: in the case of LRGB composition it is necessary to retouch the curves by hand so that the brightness profiles are as similar as possible, otherwise more or less serious chromatic defects arise.

For example, if the L is too bright compared to RGB you will have a desaturation of the image, obtaining washed-out colors that are often difficult to recover later; conversely, if the L is too dark you will have oversaturation and, above all, the onset of “ringing” phenomena around the stars (strongly colored halos but darker than the sky background).

In the case of IRGB composition, instead, it is sufficient to apply to the L a linear transformation so as to make it coincide with the I channel of the RGB image.

This apparently complicated operation can be done by PixInsight automatically thanks to the LinearFit Process.

The first step is to extract the I channel from the RGB image:

• Open the ChannelExtraction process
• Select the HSI color space
• Deselect the H and S boxes (we do not need the hue and saturation channels)
• Drag the blue triangle of the process onto the RGB image

We will then obtain a new black-and-white image with the same name as the RGB image but with the suffix _I: this image constitutes the reference for adapting the L image.

• Open the LinearFit process
• Select as the reference image the I channel of the RGB image
• Drag the blue triangle of the process onto the L image

In this way PixInsight will compare the two images pixel by pixel (this is why they need to be aligned) and will calculate the best linear transformation that adapts the L image to the I image

The I image, on the left, and the L image BEFORE LinearFit with the same STF applied: it can be seen that the luminance is much brighter than the I

the same images AFTER LinearFit with the same STF applied: now the two images are comparable in terms of brightness

After adapting the luminance to the I channel, the time has come to perform the IRGB merge; the operation is very simple: through ChannelCombination the I channel of the RGB image is replaced with the new L.

• Open the ChannelCombination process
• Select the HSI color space
• Deselect the H and S boxes (the channels will remain unchanged)
• Load the luminance image into the I position
• Drag the small blue triangle of the process onto the RGB image

In this way the L channel, with a better signal and resolution, will completely replace the I channel, leaving H and S unaltered and thus preserving the color.

here are some detail images

Linear RGB image	Linear L image
Linear IRGB Composition
Move the mouse over the image to see the effect of applying the L channel onto RGB

And here is the final image after the complete processing, click for the full-resolution version