In a vibrant, busy field of Hydrogen and Oxygen glow, a column of dust rises through the center. Three globes dominate but the lower right globe appears different.  Structurally, it resembles IC443 (The Jellyfish)  which is a supernova remnant.  This site ( https://theskylive.com/sky/deepsky/ngc6334-object) calls 6334 a supernova remnant, but specifically identifies the globe on the lower left as the SNR.  However, other sites (https://astro.i-net.hu/content/cats-paw-nebula-ngc-6334  and https://www.si.edu/newsdesk/releases/cat-s-paw-nebula-littered-baby-stars) dismiss the idea that this is an SNR.  

Intense star creation is occurring here.  It is well described here  https://www.si.edu/newsdesk/releases/cat-s-paw-nebula-littered-baby-stars:
NGC 6334 is a realm of extremes. The nebula contains about 200,000 suns’ worth of material that is coalescing to form new stars, some with up to 30 to 40 times as much mass as our Sun. It houses tens of thousands of recently formed stars, more than 2,000 of which are extremely young and still trapped inside their dusty cocoons. Most of these stars are forming in clusters where the stars are spaced up to a thousand times closer than the stars in the Sun’s neighborhood.The cause of the baby boom in NGC 6334 isn’t clear. Two processes often suggested to trigger bursts of star formation are blast waves from a nearby supernova explosion, or molecular cloud collisions when galaxies smash together. Neither is the case here.That mystery is one reason why astronomers are interested in NGC 6334. Rapid star formation is often seen in luminous starburst galaxies (like the Antennae galaxies for example). Because NGC 6334 is nearby, astronomers can probe it in much greater detail, even down to counting the numbers of individual stars of various types and ages.
Data capture occurred over the course of about a month.  Much of the data captured was dismissed due to high FWHMs.   After dismissing all subs over 2", I have the three channel masters come in between 1.81" and 1.84".  Ha and SII were strong but the OIII is quite weak outside the bright centers of the globes. 

I made several attempts at SHO combination in PI but I could not get the color balance and noise control I was after.  So I went really old school and combined stretched masters in Photoshop using Don Goldman's  clipping mask method outlined here: https://www.cedic.at/arc/c15/dwn/CEDIC15_W05_Don_Goldman.pdf  I always enjoy processing SHO data with this method because it gives me control over so many attributes at once and allows me to immediately see the results.

EDIT:

I tried something a bit different for the pseudo-luminance.  Because the OIII is weak and the SII needs a good stretch, I wanted to use a controlled pseudo-lum rather than rely on the color mix for the luminance information.  To have the minimum of noise included from the O and S, I started with the Ha and added the O and S less their background values.  This avoided some of the noisiest data.  So the PixelMath formula was like this:

Ha + S - med(S) + O - med(O)

This was done in linear space with starless images so summing these was fine and did not risk saturation.  I like this much better than resorting to a Ha only luminance which loses much of the structure of the object.