Issue with LRGB Combination.

[sfacchin]

             Hello to all,
some one can help me, please? I have a problem with the LRGB combination tool, the resulting image is a disaster. I have four separate images (L, R, G and B)  acquired with a W&B camera, and I have to combine them in order to obtaine a colour image. Before to use the above tool, I use the LinearFit tool, in order to normalizze the images. Then I use the LRGB combination tool, but the result is no good, also changing the Lightness and Saturation parameters. I tried to use different images as reference (L and G) but not difference in result. Using others SW the resulting image is very good,  whitout no problems.

Configuration:
PI 1.8 RC7, just bued!

Thanks in advance to who will help me.
Best regards,
Sergio

[StefanG]

Hi Sergio,

could you describe in more detail what you have done. Otherwise one can just guess what went wrong.
There are some excellent videos by Vicent Peris and Harry Page
http://pixinsight.com/videos/NGC1808LRGB-vperis/en.html
http://www.harrysastroshed.com/lrgb.HTML
that describe how it is done. Basically, the recommended way is to first combine the R,G and B channel into an RGB image and perform dynamic background extraction, background neutralization and color calibration on it. Then you would stretch the RGB and the L in the same way and perform HDRCompression in order to avoid problems with pixel saturation during LRGB combination. This might be one of the issues you have. Then you perform the LRGB combination with either the LRGB tools or the channel combination tool.

Stefan

[GJL]

Hello Sergio,

Stefan has described the way, RGB and then LRGB Combination. LRGB Combination Tool may be better, than ChannelCombination, you´ll get better colours. Try Lightness 0.500 and Saturation ~ 0.170, less Sat. value will result in more colour.

LRGBComb.:  Check L, uncheck R,G,B and drag the triangle to RGB image.

Regards   Gerhard

[sfacchin]

           Thanks,
Stefan and  Gerhard for your prompt replay.
During these days I ran some tests with PixelInsight (PI). My first goal was to get a LRGB image in linear space. For this I followed the procedure from this video, (http://dl.dropbox.com/u/57910417/Video03_Kombination_OIIIHaLRGB_END.wmv) but the result was different from that expected. In the video, reference is made to the previous version of PI, perhaps, the Channelcombination works in a different way.
In any case, to obtain an image LRGB in the linear space RBG just use the instrument Pixelmath.

Seeing the video you suggested (Stefan), I realized that the tool LRGBcombination, works in the CIE L*a*b* space. This means that before combining the two images they must be stretch. In the video, they use the stretch obtained by the tool ScrenTranfertFunction used in autostretch mode. So it is very easy to obtain the same level of stretch for the two images. However, the stretch thus obtained is very high and the final image shows a lot of noise. Then serves further work of filtering to reduce noise.
I tried to follow the procedure indicated in the video, but without using the automatic stretch. I used the script MaskedStretch for both images and subsequently stretch through HistogramTransformation for both images. I made sure that the stretch was approximately the same for the two images and the apply the LRGBCombination tool. The result is better since I have less noise while having a good effect produced by luminance. I'm still experimenting, I'll let you know.

Thanks and regards,
Sergio (Venice Italy)

[StefanG]

Hi Sergio,

Perhaps one more hint for your experimentation: The defaults of the autostretch function in builds < 1054 are quite strong and can reveal a lot of noise in the images, as you pointed out. You can customize its behaviour however, by pressing <ctrl> and clicking on the autostretch button (radioactivity symbol) in STRF. Here you can reduce the desired median to around ~0.125 perhaps and reduce also the amount of clipping. That would give you a reproducable but less harsh stretching.
Of course using the approach that Gerald demonstated very nicely in his video (Thank you very much Gerald!!!) is also possible.
Best regards,

Stefan