LRGB Saturation Study 
June 16, 2002

LRGB combination technique has become the preferred method for creating color CCD images. This technique allows the imager to shorten the exposure time, while gaining in S/N. 

There have been numerous reports that the LRGB process reduces the vibrancy and contrast of colors in the image. This is most evident in the star colors of many images processed this way. Very often the stars are left mostly white with very little color saturation. The purpose of this study is to analyze and quantify this effect, and to possibly  suggest a better way to do LRGB.

1. How the study was conducted:

1.1 Measuring Saturation

Color saturation is defined as the amount of color per unit of brightness. To measure color saturation in Photoshop I used the color sampler tool set to display the Hue, Saturation and Brightness components (HSB).

1.2 LRGB Combination methods

I investigated two ways of creating the  LRGB composite in Photoshop:

Method 1 involves using Photoshop layers by adding a new layer to the RGB image containing the Luminance (grayscale) frame. The combine method is set to Luminance at 100%.

Method 2 involves conversion of the image to Lab color space. In Lab, the luminance channel is then selected, and the luminance frame is pasted directly into it. The resulting image is then converted back to RGB.

Method 3 is a method I developed to preserve the saturation of the original RGB image, while adjusting the brightness values based on the Luminance frame. I'm calling this the "Color Ratios LRGB" method, described here.

1.3 Images

The images used in this study were created for the study. 

The RGB image contains three intersecting color circles, each at half the maximum display value (128). The red circle has the RGB value of 128,0,0,  the green circle is 0,128,0 and the blue circle is 0,0,128.

 
  Original RGB

The grayscale image was created by converting the original RGB image into grayscale using the Grayscale command of Photoshop. An additional image was created by multiplying the brightness of the grayscale image by 2.


Grayscale


Grayscale x2

1.4 Process Description

The study involved creating LRGB images by combining the above images using the Luminance layering, the Lab luminance replacement, and the new method. The saturation of each of the resulting images was then measured and catalogued.

2. Results

The saturation in the original RGB image is 100% in each of the primary color circles. HSB value shown below was measured in the green circle.

Combine Method LRGB with Grayscale LRGB with Grayscale x2

Luminance Layering


89% 
HSB: 120, 89%, 50%


28%
HSB: 120, 28%, 100%

Lab L channel replacement


100%
HSB: 120, 100%, 46%


61%
HSB: 116, 61%, 100%

Color Ratios LRGB


100%
HSB: 120, 100%, 76%


100%
HSB: 120, 100%, 91%

3. Conclusions

From the above study, it appears that the standard LRGB methods in Photoshop can result in a significant loss of color saturation. Of the two combine methods, the Luminance Layering produces by far worse results than the Lab luminance replacement method. But, both methods reduce color saturation if the brightness of the Luminance frame is greater than the brightness of the original RGB image. This may explain why the stars desaturate a lot in LRGB practice, as they are probably the brightest object in the luminance image. The Lab luminance replacement method seems to shift the color hue slightly, but I'm not sure that this shift is significant or noticeable.

The results indicate that one possible way to avoid the desaturation in LRGB is to lower the brightness of the luminance frame to be below or to match more closely the brightness of the RGB image. When done this way the Lab luminance replacement method will preserve  most of the saturation of the original RGB. 

The Color Ratios method shows promise. It seems to keep the saturation constant even while increasing the brightness levels. Unfortunately, it does not keep the brightness level constant, but that's something that can be easily corrected with level/brightness controls. To see a description of the method, click here.

I welcome all comments and suggestions sent to e-mail.

Copyright 2001,02 Paul Kanevsky