This is meant to be a part of a set image for the M95/M96 galaxy group. The image for M95 is here.  The images for the two galaxies are composed of similar amount of exposures and are processed similarly.  Both are LHaRGB compositions.

Previously when I imaged the two galaxies with smaller telescopes, I got the impression that the two sister galaxies are just copies of each others. They have similar sizes, brightness, and shapes (both with double rings plus a central bar).  They both look like a Θ.  However, after I imaged them with the 20" scope under <1.5" seeing, I no longer feel they are similar.

The slightly smaller one of the two, M95, has a clear bar at the center, and has spiral filaments of dust lanes across the entire disk. Star-forming regions (the red H-alpha dots) also spread over the disk, and mainly follow the spirals. Clearly this galaxy has a structure driven by the central bar and the spiral density wave propagating through the disk. At the center of its bar, there is a mini spiral and a nuclear star-forming ring. What appears to be an outer ring is actually two tightly wounded spiral arms. In short, this is a spectacular spiral galaxy without a strong sign of interaction and perturbation by another galaxy.

M96, the larger one of the two, on the other hand, is clearly undergoing galaxy interaction. It has a large/bright central disk with spiral dust lanes on it. I am not sure if it has a bar at all. The outer ring is either a ring or a one-arm spiral. Such a structure is usually induced by flying-by of a satellite galaxy. And indeed, there is a compact elliptical (PGC83335) at the lower-right edge of the ring and it shows signs of interaction with M96. The ring and the central disk contain most of the dust and star-forming regions, and it is quite empty between them (you can easily see through it and find a handful of background galaxies).

I hope you enjoy the images.


Rev.B: a comparison of RGB and LRGB under same total integration time and same minimal processing

Recently there was a forum post claiming that adding L to RGB data degrades the quality. In response to that, I used partial data for this image to create RGB and LRGB versions that contain identical amount of total integration time (208 minutes) to demonstrate the effectiveness of the LRGB composition.

Procedure: I removed exposures with very good and poor seeing, so the combined R,G,B and L images should all have similar resolution. At the end, I have the following:
B: 5-minute x16
G: 4-minute x16
R: 4-minute x16
L: 5-minute x21
The above give us an RGB color image that contains 208 minutes of total integration, and an L image that contains 105 minutes of integration. Then I try to create another RGB image that only contains 104 minutes of integration (half of total). I did this by alternatively selecting every other of the RGB exposures, rather than selecting the first half or the second half. This way, whatever which interlaced half of the RGB exposures is selected, the quality should be fairly similar to the other half and to the full set. 

The 208 minutes of RGB, 104 minutes of RGB, and the 105 minutes of L were stacked and processed in identical ways in PI. The 104-min RGB image was linearly fitted to the 208-min one to ensure identical brightness, contrast, and color balance of the two RGB images. Then the 104-min RGB and 105-min L were combined in Photoshop to create a 209-min worth of LRGB image. The LRGB composition was done in linear space. Then all the images were given nearly identical amount of screen stretch on contrast and saturation (otherwise the linear images would just look black). They are "nearly" identical rather than entirely identical because the linear fit procedure did not reach identical brightness and color likely because of the imbalanced noise in the two images, and also because images with stronger noise can appear brighter to our eyes even though the averaged brightness is the same. So there is a tiny difference in the screen stretch parameters to compensate this small difference.

The left column and middle column show images without any further processing, just stacking, linear fitting, and screen stretch. So you are viewing the data's most original form. You can already see that with nearly identical amount of time spending on imaging (208 minutes vs 209 minutes), the LRGB one gives you a smoother looking and better defined details on faint features that are closer to the noise floor.  That's the help from the L.  If your eyes are well trained enough, you can still see strong color noise in the LRGB image, since the quality of the color is only 104-min worth.  So this is less ideal, comparing to having 208 minutes of RGB data. However, the beauty of LRGB composition is that our eyes are more sensitive to fine image structures in luminance, and less sensitive to fine structures in color. This allows as to apply a color-only noise reduction without changing the overall sharpness of the image (under human eyes). (The NR applied here is very mild. One can actually apply a much more aggressive NR to the color without damaging the sharpness.) This is the upper-right panel.

Conclusion: I think most people can agree that both the LRGB without noise reduction and LRGB with color-only noise reduction look better than pure RGB image of the same integration time. so unless you have infinitely amount of time to spend, it will be a good thing to do to spend some time (not necessarily half) on L. At the same time, I want to admit that the difference between the RGB one and LRGB one can be subtle to some people. But to those who push the data to the limit, this subtle difference matters.

ps, the above conclusion only applies to continuum objects, like galaxies, reflection nebulas, stars etc. It doesn't apply to emission line objects where an L filter doesn't necessarily collect much more signal than an individual R, G, or B filter or a narrowband filter. 

ps2, you may say: "If you are allowed to do NR on the LRGB image in this comparison, why not doing NR to the RGB one as well (to be fair)"?  Well, actually you can of course do the same NR on the pure RGB image if you insist a fair comparison.  Indeed, I tried it.  Unfortunately NR on the RGB image is less effective.  This is because its luminance noise is correlated with color noise (coming from the same data).  If you just do a color-only NR (the same amount as what was done on the LRGB one), it doesn't reduce the feel of noisiness much, as the noise in its luminance is still there.  And if you allow the luminance to also undergone the same amount of NR, the sharpness gets some hit.  And if you use some fancy NR routine that can preserve sharpness, then for the sake of fair comparison, the LRGB one can also employ such advanced NR for its L component.  So no matter how you do it, as long as you try to stay reasonable and fair, the LRGB one will win (until your NR becomes so strong that it hides all the meaningful differences). 

ps3, There are definitely other ways to make an RGB vs LRGB comparison. But in my mind, if your goal is to compare the superiorness of the two techniques, same total integration time, same data quality, and same minimum to no processing should be the round rule for the comparison.

ps4, The reality is that our seeing rarely stay still and it changes. If you can measure seeing in real time and adjust your image plan accordingly, the difference between RGB and LRGB will be even bigger. You can achieve even sharper images if you only do L under better seeing, and reserve RGB to poor seeing. That's usually what I do.