An iTelescope Article by Dr. John Ebersole

"If you enjoy photographing nebulae, I hope you will try narrowband imaging.  The results are worth the small amount of extra effort."

For many of us, our first view of a narrowband image was the 1995 release of the “Pillars of Creation” from the Hubble Space Telescope (HST) (Fig. 1). 

It was significantly different from the typical views of red emission nebulae, when RGB filters are used.  Not only were the colors different, but there was remarkable color contrast and a 3-D effect unusual for traditional color images.  This image was taken with emission line filters for ionized Hydrogen alpha (Ha), Sulfur II (SII), and Oxygen III (OIII).  These three grayscale images were combined into an RGB color image using the now classic, wavelength-ordered (longest to shortest), HST palette, wherein SII is portrayed in red, Ha in green, and OIII in blue.

Figure 1 – Original “Pillars of Creation” (left) and an enlargement of a 2016 T24 image, both in HST palette.

Spectra from nebulae show specific emission lines, and proper filters can reveal a nebula’s ionization structure by passing only a “narrow band” of wavelengths around these lines (see Fig. 2). 


Moonlight and various forms of light pollution are rather broadband, so little of their energy passes through these same narrowband filters.  As a result, greater contrast is achieved.  The practical benefit is that narrowband imaging can be done under conditions of moonlight or city-generated light pollution, when RGB imaging would be confounded.

The most common emission line found in nebulae is Hydrogen alpha at ~656nm.  It provides the “red” component of an RGB nebular image.  The correlate Ha filter typically captures the most photons and the most nebular structure of this filter type. The Ha image is as a result often considerably brighter than those of other emission lines, so adjustments commonly need to be made in processing.  Nearby in wavelength is the emission line doublet, Sulfur II, at ~672nm.  Typical red filters pass both Ha and SII so that the contrast between them and any structural differences in their images are lost.  The SII emission is considerably less intense than that of Ha in most cases, so exposures that are longer, more in number, and/or binned 2x2 are usually required.  The other important emission line is the doublet, OIII, at ~498nm.  It provides the blue-green color of many planetary nebulae in RGB images.  It too is often less intense than Ha and requires additional exposure or 2x2 binning.

There are several narrowband palettes, and most use images of these three emission lines in a variety of orders, as RGB components of a resultant color image.  The HST palette noted above is by far the most commonly used.   Another is the Canada France Hawaii Telescope (CFHT) palette, in which the Ha emission is perhaps more naturally displayed in red and OIII in green, but SII is relegated to an unusual blue portrayal (see Fig. 3).

In most instances, the HST palette tends to produce the most dramatic and aesthetically pleasing images because of the structural relationships of these emissions in many nebulae.   OIII is often diffuse, and in curved or ovoid nebulae it tends to be concentrated in the concave side or center of the nebula, respectively.  SII, on the other hand, is typically found at the outer edge or convex side of the nebula, adjacent to the Ha.


Given these juxtapositions, a natural spectral gradient from blue through green, yellow to red is often achieved across the nebula from the concave to convex side or from center to outer edge.   The yellow to red to brown gradient across the Ha and SII of the major structural components of the nebula can give it shading that results a 3-D effect, not seen in RGB or other narrowband palettes (see Fig. 4)


These generalizations do not apply to all emission nebulae.  This is particularly true for supernova remnants (SNRs), where bright filaments of each emission line can fall in no specific pattern or order.  This too results in a modification of processing described below.  Also note that narrowband imaging does not improve reflection nebulae that tend to possess mostly broadband light.

Figure 3 – The same Ha, SII, and OIII frames were combined at the same weightings in HST palette (left) and CFHT palette (right) of IC 4628, the Prawn Nebula. Note that the wavelength-ordered HST palette provides a more pleasing and structured image, particularly of the intersection of Ha and SII gradients at the convexity of the nebula. (2016 T32 image.)

Figure 4 – Narrowband images illustrating the 3-D effect of the HST palette. Outer Eta Carina nebulae from T31 (left); widefield Cygnus nebulae from T14 (right).


There is no set exposure duration or exposure number that works best for all emission nebulae.  If you have no restrictions regarding time or expense, then the more total exposure for each emission line the better.  However, most of us have one or both of these restrictions, so some compromise needs to be reached.

For most emission nebulae, the Ha emission will carry the majority of structural information.  Accordingly, it should be given a large share of the effort and be shot at 1x1 binning to get the best resolution.  Ha can be obtained safely at exposures of 600sec, which is the maximum for most iT scopes.  If there are tracking issues, reduce the exposure to 300sec.  I have found that 120 min total Ha provides good results for most medium bright nebulae.  Bright nebulae, such as M8 or Eta Carina, may need less exposure; dim nebulae will likely need more.  

Remember, you will achieve full resolution of the brightest parts of the nebula with only a few stacked images.  Additional imaging simply brings the dim parts of the nebula out of the backgound noise.

Here is where you can compromise and/or use filtering.

The OIII and SII emissions are typically dimmer than that of Ha.  Also they are less likely to carry unique structural information.  Accordingly, I commonly image these at 2x2 binning to enhance photon capture by a factor of 4.  600 second exposures and a total duration of 1 to 1.5 hours at this binning will usually provide sufficient data for a reasonable image.  Obviously, if there is non-redundant structural information in these emissions, as is usually the case with SNRs or very bright nebulae, they should also be imaged at 1x1 binning (see Fig. 5).  

Some of the finest narrowband images with velvet-smooth backgrounds are obtained by extremely long total exposures, e.g. 20-30 hours.  Improvement over a typical 4-5 hour image is asymptotic , however, with progressively diminishing returns.  You have to decide with what you will be satisfied.

Figure 5 – In supernova remnants, such as the Vela SNR (left), unique structures are seen in OIII and SII emissions, in addition to that of Ha. Also in bright nebulae, such as Eta Carina (right), OIII and SII emissions may be as intense as Ha. In both cases, imaging should be done at 1x1 binning with all narrowband filters. (The bi-color Vela SNR image (OIII in blue, Ha in red) from T8; HST palette image of the Eta Carina core from T27)

Remember, narrowband imaging can be done in moonlight.  As long as the moon is at least 60 degrees from your target, you can image in Ha and SII, even when the moon is full.  For best results, the moon should be at least 90 degrees from your target for OIII imaging.  I typically shoot OIII when the moon is down and Ha, SII when the moon is full and the  iT discount is 50%. 


Ha, OII, and SII images of good quality should be stacked separately by whatever program you use.  I use Maxim DL.  (Some may consider this “old school”, but then again so is the operator.)  Prior to stacking, I typically eliminate hot and dead pixels and column defects.  I usually use an SD mask routine for the stacking. The resultant image data should be stretched linearly, so that most of the 16-bit range of brightness values are used.  I then bring each stacked image into Photoshop for additional processing.


The next important step is “curves” stretching to enhance the lower brightness nebulosity as opposed to the bright stars.  This is done in successive steps with a readjustment of the black level with each stretch.   Some would advise that the goal is to get the brightness histograms of the three stretched emission images to look similar.  This is fine, but a simpler goal is to “curves” stretch each image successively, until the noise in dim parts of the image becomes objectionable.  At this point you can go no further with the data you have.  To visualize better these dim areas, you have to have more exposure.  Alternatively, as mentioned above, you can filter the image to reduce noise.  Such noise reduction filtering will undoubtedly also reduce the sharpness of detail in the bright parts of the nebula.  An important tip for processing of all kinds is to use the “lasso” or other similar tool with feathering to outline the area where you want modification and to exclude areas that do not need change.  This usually means outlining only the dim areas for noise reduction, etc.

A variety of additional tweaks can be done before combining the images.  These include enhancing the contrast, reducing the size of stars, and true unsharp masking.  Given the character of the initial images and the amount of stretching, stars are usually larger in resultant OIII and in particular SII images.   If combined without reduction, stars will have a red or blue rim.  There are a number of techniques for star reduction.  I use Carboni Astronomy Tools Actions.  Additional contrast can sometimes be achieved with true unsharp masking.  I recommend the technique proposed by Jerry Lodriguss (www.astropix.com).

Bright nebulae offer as much or more of a problem than do dim ones.  The range of brightness is so great that it cannot be contained in one standard image.  For example, the nebular core may be totally “burned out” when stretching the image to the point where the outer nebulosity is apparent.  There are several ways to handle this problem.  The easiest is a processing tool called digital development processing or DDP.  By adjusting various thresholds, the large dynamic range of the image is compressed.  I favor a somewhat labor intensive approach because it yields better contrast.  This involves adding multiple layers of progressively less stretched and manually outlined image segments until you achieve an image clearly showing all brightness levels (see Fig. 6).

Figure 6 – Nebulae, such as M8, have a large dynamic range from the bright core (left) to the dim periphery. Stretching to visualize the latter will ”burn-out” the core (middle). Several methods are available to compress this range in order to see details of both bright and dim structures (right). (T24 Ha images)

Once the final adjustments have been made to all three emission line images, they need to be combined in a narrowband palette to form an RGB color image (Fig. 7).  I use the classic HST palette.  This step I perform back in Maxim DL.

Despite maximal stretching of OIII and SII, it is likely that these images will have to be weighted in comparison to Ha for the color combine.   Combining with a 1:1:1 ratio will likely produce an overall green image, given the dominance of Ha, even after stretching.  After aligning the Ha, SII, and OIII images, I proceed on a trial and error basis combining the images with variable additional weighting of the OIII and SII.  For example a combine ratio of 2:1:1.5 for SII:Ha:OIII, may produce the most pleasing result.  This part is not science; it is art.  The best combination image(s) is brought back into PS for final adjustments (see Fig. 8).

Figure 7 – Ha, OIII, and SII images (left to right) of NGC 6188 ready to be color combined into an HST RGB image. (T33 image)

When you weight SII and OIII more than Ha, the stars in the resultant image will no longer be white, but probably magenta (the OIII blue and SII red color combinant).   Reducing the magenta saturation in PS, for example, will eliminate this problem.  Some prefer to add RGB stars, in a separate step.  Additional color balance adjustments may be needed.  Regional stretching or color balance adjustments can be performed using the lasso tool, as cited above.  Some degree of sharpening via a variety of programs may improve the image further.  Reducing the size of the final presentation image from that of full resolution will also tend to smooth the result.  Finally, you are done!

Figure 8 – Combining SII, Ha, and OIII images of NGC 6188 at weighting ratios of 1:1:1 (left), 1.5:1:1.5 (center) and 2:1:1.7 (right). Final adjustments have been made in magenta desaturation, color balance, contrast, star size, sharpening, and cropping in the right most, final image. (T33 image)


If you enjoy photographing nebulae, I hope you will try narrowband imaging.  The results are worth the small amount of extra effort.  iTelescope.Net offer an easy way to try narrowband imaging at multiple fields of view (Fig. 9).  


  • Start with a medium bright emission nebula that does not require dynamic compression.  
  • Use posted and published images to judge your results.  Have fun!  
  • But be aware - narrowband imaging can become addictive.  I should know!


Figure 9 – HST palette narrowband images of NGC 6357 at multiple scales from a widefield T12 view to nebular core imaging with T27.


I am indebted to Richard Crisp (www.narrowbandimaging.com), pioneer of amateur narrowband imaging, for my initial interest in and subsequent development of skills in the technique.

All images except the HST Pillars of Creation have been acquired by the author with iTelescope.Net equipment.

Additional narrowband images can be seen at the author’s website :  http://jebersol.zenfolio.com/


Thank you John.

Many of John's APOD winning images can also be viewed in the iTelescope APOD Gallery