Planetary Nebulae
Planetary Nebulae
Planetary Nebulae
     HOME          TABLES I,II,II,IR          DOCUMENTATION          DATA GATHERING                                      
Advanced image processing techniques to detect and identify faint PN candidates
  June 13th 2023 - Peter Bresseler     
The following article describes the use of a advanced image processing methode to improve the visibility of faint objects, with a primary focus on planetary nebulae (PN), based on optical surveys. The procedure removes stars from the image, leaving only the background and the object to be analyzed. By separating the stars from the images, it is possible to control and manipulate the data more consciously and to concentrate on the essential in the image processing, Object structures and their morphology are more clearly identified, artifacts are excluded.

Key words: optical surveys - advanced image processing - planetary nebulae identification

Astronomers such as Georg Abell searched visually for planetary nebulae on the POSS-I plates in the 1950s and 1960s. Today, the number of available surveys has increased considerably but the search methods do not differ significantly from those used in the past. The inspection methods are based on the visual examination of digital sky surveys in combination with optical and non-optical wavelengths. Once a structure or object of interest has been identified in one or more surveys at the same time, the research process begins with the analysis and inspection of the image(s) and study available data and informations.

Survey inspection
Today the search for PN is typically based on visual inspection mainly from digitized optical surveys. Since H-alpha is one of the most significant and importend emission lines, it is obvious to use the corresponding Halpha-surveys. For the northern hemisphere IPHAS – Isaac Newton telescope Photometric Halpha survey [1] is suitable, for the southern hemisphere SHS – SuperCOSMOS Halpha survey [2]. But also colored surveysyes like PanSTARRS [3] have their relevance, as we will see. The surveys can be accessed directly through the corresponding web-sites or through various interactive webportals such as [4], Aladin Sky Atlas [5] or ESASKY [6].

Candidates and confirmed PN
Many of the large and bright PN are already registered. Citizen science projects in which amateurs work with astronomers, have a high value. Planetary nebulae candidates were uncovered by amateur teams led by Matthias Kronberger (Kronberger et al. 2010), which is no longer active and today by the very active team led by Pascal Le Dû (Acker & Le Dû 2014).
Regarding optical surveys, the previously undetected candidates are faint interstellar structures, dust arcs of low signal strength that stand out only faintly from the background of the surveys. These are usually very faint H? signatures or have weak greenish structures in PanSTARRS colors. Remaining searched objects are extremely faint and mostly of small size, only a few arcseconds in size.
Potentially, interesting candidates may have a diffuse circular, the visible appearance of a ring (complete or partial) or a disk or a bubble, but some may have a bipolar morphology or be more atypical. Others have only a stellar appearance. A candidate of interest can be identified by three criteria: its morphology, its signal in the underlying wavelengths and its color. (Le Dû 2016).

Visual inspection
Visual inspection focuses on isolated objects that are not part of a larger structure. This differentiation is important, otherwise the candidate may not be a true candidate but part of a larger HII region. But first that must be detected. As a rule, the objects hardly differ from the background, often they merge with it. Therefore, for better identification, a histogram function is used to change the pixel intensity on the x-axis so that the background boundary is reached. Today, a histogram function for changing the brightness distribution is part of every image application or function of a web portals.

Image processing methods
If one or more characteristics match at the same time of a potential candidate, such as morphology, signal strength, or color, more in-depth and accurate image processing usually takes place. This includes classic mathematical functions for image processing. Basically, the first analysis is to determine the morphology of a faint PN candidate.

For better identification, typically a subtraction or quotient imaging has been applied to digitized photographic survey plates and films.(Frew & Parker 2010). These techniques have proven to be very successful today.
The difference method is used to determine the morphology of possible candidates and generates an expressive result.

Fig1: (left) IPHAS Halpha-band of PN Ou 3 (PN G 059.2 + 01.0) with northeast to upper left. (middle) IPHAS R-band of Ou 3. (right) IPHAS Halpha/R-band difference image of Ou 3. Image subtraction subtracts the IPHAS R-band from the IPHAS Halpha-band and produces a difference image. The PN OU 3 is clearly visible and comes out well.
Remove stars
With the increasing use of digital cameras in astrophotography, more advanced image processing routines and techniques have been developed. As part of image processing to create pretty images of deep-sky objects, astrophotographers sometimes use a method to visualize fine structures of deep-sky objects by removing stars from the present image, stretching the starless image considerably, and then adding the stars again.
This procedure could be applied in a modified form in the identification of a faint PN candidate and it could be used to advantage here.

However, it is not always just a matter of clearer identification, sometimes it is also necessary to recognize and exclude artifacts. In both cases, identification or exclusion of artifacts, the following adjusted procedure can be applied:
        1. Remove the stars from the image, you get a new image without stars; the background is homogeneous, the PN candidate is more clearly visible
        2. Increase contrast and pixel intensity of the starless image, e.g. the brightness ratio with the use of a histogram, this makes the PN candidate of the starless image stand out better, the background darker. At the same time artifacts can be detected. With a starless image, it is easier to stretch the faint structure without fear of blowing out the stars.
        3. Blend the starless image with the original image, e.g. by adding or multiplying the starless image with the original image.

Annotation: Step 3 can be skipped, since only the morphology is to be visualized.
Fig2: (left) PanSTARRS I-R-G band of PN LDu 13 (PNG 114.4+00.0) with northeast to upper left. (right) PanSTARRS I-R-G band of PN LDu 13 after remove stars. Very bright stars remain, could be manually merged with background. After remove stars the structurstructure is better recognized and the PN LDu 13 comes out well.
By separating the stars from the images, you can more consciously control and manipulate the data overall. It is possible that the background will be increased in addition to the PN Candidate, but not stars that are blown up. The advantage, the faint contrasts are raised, possible PN candidates become more visible, without the star being affected.
Star removal techniques are not new and are implemented either as a function or as a plugin in commercial image processing programs such as Pixinsight7 or Adobe Photoshop8. Also we find the possibility to integrate such a function in free image processing programs, SIRL9 is to be mentioned here.
Furthermore, individual, classic image processing functions of the mentioned programs can also be combined to achieve the goal of removing stars. On the web there are tutorials for the most common image processing programs to remove stars from images.
Fig3: (left) IPHAS Halpha-band of PN LDu 13 (PNG 114.4+00.0) with northeast to upper left. (right) IPHAS Halpha-band of PN LDu 13 after remove stars. After removing stars and using a histogram, the structure is better recognized and PN LDu 13 comes out more clear.
StarNet++10 is only used for the removing of stars from a image. The image can be an RGB image or just monochrome. As a prerequisite, the files must be saved in a 16-bit format. Typically, screen-shots are used from digital surveys which are saved as 16-bit files.
After starting Starnet++ the image file has to be selected, also a destination file has to be specified, click, run, done.
Starnet++ analyzes the images and implicitly generates a synthetic star mask of the image. The synthetic star mask is subtracted from the original and the gaps are filled with the surrounding background, so a clean and smooth background is created. Starnet++ detects stars very well in an image, but for very bright stars artifacts may remain after subtraction.
The background without stars can now be stretched without affecting the stars. The central star of PN (CSPN) are also removed with this method. But this is not critical, because they are present in the original and can be blended or added back into the original at the end of processing. We want to stretch the sky background maximal to detect a faint PN, for this Starnet++ is very helpful. [10]

There are some limitation which is derived from the usage with starnet++ respectively the image processing method described here.
• Inspections at non-optical wavelengths such as in the ultraviolet (UV), MIR, and radio ranges are difficult or impossible to process because the signals are roughly
• Star shaped small objects do not have halos or extended structures and only a few arc seconds in size. In the course of the image subtraction small objects are taken away, thus also potential candidates.
• If the remove stars function is applied to PanSTARRS color images, all stars are removed, possibly also the CSPN.

The usage of starnet++ or the image processing method described here can be used in all cases where image analysis is required to determine an object, it’s morphology or to exclude artifacts. These include supernova remnants, extended HII regions and, as shown here, the identification of PN candidates. The method has its limitations as shown, but it is particularly well suited to work out the morphology more clearly. Properly applied the using advanced image processing methods it is possible to increase significantly the power of determining a PN candidate.

- Jacoby, G. H., Kronberger, M., Patchick, D., Teutsch, P., Saloranta, J., Howell, M., et al. (2010). Searching for Faint Planetary Nebulae Using the Digital Sky Survey. Publ. Astron. Soc. Aust. 27, 156–165. doi:10.1071/AS09025
- Acker, A., and Le Dû, P. (2014). Nébuleuses planétaires: joyaux de l’astrophotographie. L’Astronomie 128, 40.
- Le Dû, P., 2016, Comment Découvrir de Nouvelles Nébuleuses, L´Astronomie, Vol. 130 | 91 |35
- Frew, D. J., & Parker, Q. A. 2010, Publications of the Astronomical Society of Australia, 27, 129–148
Sources and links
Planetary Nebulae
Respect capital letters and spaces !
Example: PN-G 262.4-01.9 or DeGaPe 32
  About the article...
Date: June 13th 2023
Author: Peter Bresseler
Article id: 56
All the articles: click here
Associated articles:
- Pa30, SN1181 et le travail des amateurs: dossier c...
- Une collaboration ProAm récompensée...
Planetary Nebulae .net
Contacts: - in charge PN and applications: Pascal Le Dû (
- webmaster: Thomas Petit (
© Planetary Nebulae .net - All rights reserved
The Planetary Nebulae .net website offers all the useful and practical information about the planetary nebulae (candidates), how to apply for a new discovery and how to confirm the candidates. If you would like to contribute to the website, please visit the Data Gathering page and contact us.
Contributors: Pascal Le Dû - Thomas Petit - Olivier Garde - Nicolas Petit - Trygve Prestgard - Jean-Baptiste Feldmann - Lionel Mulato - Sakib Rasool - Dana Patchick - Xavier Strottner - Jean-Baptiste Auroux