This is a reprocess of the Prawn Nebula, aiming to show as carefully as possible the relative topographical distributions of H-alpha (yellow) and SII (blue).
We were concerned about artefacts in the distributions due to a bad zero point or to differential nonlinear stretches. Consequently, we set the zero point for each channel meticulously to the foothill of the histogram, and then performed only linear stretches.
Stars were automatically whitened somewhat, in order to reduce blue SII rings which would otherwise dominate the H-alpha to SII ratio. Next, SII was linearly stretched relative to H-alpha until the grand mean brightness for each channel was the same. This could be thought of as normalizing the relative contributions of H-alpha and SII. Finally, some wavelet sharpening was applied.
A strong curvilinear stretch tends to reduce saturation and disguise differences in H-alpha vs SII distribution. The purely linear stretch helped maintain saturation.
The original image is here.
Ha 4hrs. SII 3hrs. Aspen CG16M on 20" PlaneWave. Field approx 36 min arc. Scale 0.55 sec arc/pixel.
Result: SII is relatively much stronger along the back of the prawn, that is to say the bit where the 'vein' would be. There is also a lesser, wispy, but easily visible, concentration of SII reminiscent of steam above the well-barbecued crustacean.
The astrophysical significance of the separation is unknown to us, beyond noting that SII will usually be selectively found in very thin but strongly ionized gas that contains recycled material from the depths of previous stars. Conversely H-alpha will be selectively found where the bulk stuff is.
We notice that there is no general SII 'wash' throughout the image, as might occur as an artefact of incorrectly setting the zero point and then applying a strong nonlinear curve.
Conclusion: It is important to sniff prawns carefully prior to barbecuing, as there can be sulphur along the upper surface of the back. Armed with this knowledge, the de-veined prawn can look surprisingly beautiful and different.
We thank Andy Campbell for suggesting posting this image.
Best,
Mike and Trish
the back of the Scampi nearer the tail end looks so solid and 3D 
like a cumulonimbus thunder cloud/ridge. Yes, there is definitely differentiation in this region between all three wavelengths, so to me, ideally if you take data through three filters then three colours or hues should end up in the final product and with smooth transitions between the three, or else you have kinda wasted the effort of collecting data through the third filter 
It is easy to lose the three colours without really realising it, mainly through heavy colour manipulation in processing. Having said that, the final image can still look good with a more bichromatic colour scheme, as is evidenced by the many fine Bi-Colour (usually Ha & OIII) wavelength ordered emission line images out there 
