Astrophotography in India - V

 Astrophotography in India - V

 

Stacking

As a child, I used to find NASA images hauntingly beautiful. Resplendent nebulae looked like carefully sprinkled gulal across the canvas of the universe; they were nothing short of mesmerizing. I am sure I was, but one of billions, who found inspiration from those photographs. As a grown up, I got a chance to work in labs where instead of peering up though telescopes, I was looking down the eyepiece of microscopes. The microscopic world can be just as mesmerizing as if it is a universe unto itself. The issues that astronomers face is often the same ones plaguing those who are trying to image the universe of microbes – poor signal to noise ratio, optical train effects etc. Over my years of experience with optical microscopy, I learnt important image processing techniques and basics. Little did I know then, that they would come in very handy when I started astrophotography.

Deep-sky astrophotography if often an uphill battle against noise. For a quantitative assessment, signal-to-noise ratio (SNR) is often used to assess the quality of ‘signal’. In the image panel in Figure 1, y-axis represents the intensity of measured signal and x-axis is a spatial or temporal extent. During any measurement, the measured signal usually consists of the actual sginal (say a star) along with noise. This noise could be representative of light pollution from a city or atmospheric scattering or even background noise of the universe. If there is very little or no noise, the signal can be well captured (left of Figure 1). If however, the noise is extremely strong it can distort the measurement of the underlying signal as in the rightmost panel of Figure 1. Any amateur astrophotographer can attest to the fact that raw images of telescopes are far removed from their fanciful cousins that appear in public domain. If that is the case, what goes into the making of those spectacular images? Those images are almost never ‘raw’ images obtained from telescope. Images that are obtained can undergo several hours of post-processing depending on the source and type of image. Take for example Figure 2. The background image captures in wide-field the Orion constellation. Approximately, in the mid-section of the image, M42 is clearly visible as it outshines nearby stars. The Flame nebula lies to the left of the leftmost belt star (Alnitak). From a place like Bangalore, the Flame Nebula is extremely difficult to capture due to high light pollution. The inset, which shows both the Flame and Horsehead nebulas, was captured by using an optical light filter placed in the optical train. The collected images were still not enough and had to be stacked together for a clear picture.

Figure 1: What noise can do to ‘signal’? (Left) A sample signal with no noise ((Middle) Signal with small amount of noise added to it (Right) Strong noise riding on the signal.

Figure 2:Main image was captured using a wide-field lens. 

We will discuss hardware filtering in a later essay, but here we will focus on the idea of stacking. Stacking, just as the name suggests creates a super-posed image from several sub-exposures (Figure 3). But what does this achieve? A simple act of averaging several sub-exposures is useful because it can increase SNR. The act of averaging across many sub-exposures allows random noise at a location to effectively cancel each other out, whereas the signal strength remains intact. However, the process of decreasing the noise leads to higher SNR which improves image quality. This idea is showcased in Figure 4, where the noise profile in a single sub-exposure is compared with the noise profile of a stacked image. The freeware ImageJ was used towards this end. On the left hand, you see a single sub-exposure of the Flame nebula. The image was taken using a 300 mm f/5.6 lens with an exposure of 180 seconds. The image itself is obviously very noisy; one can see both Flame and Horsehead nebulas (red) and Alnitak, but the background is obviously very noisy. To visualize the noise profile, we select a line (shown in yellow) and plot the gray-scale intensity value of the image. The flat-topped section of the curve corresponds to Alnitak, showing that the bright star has saturated the pixels at this exposure time. However, note the high-degree of noise around the star. Once these images are stacked, the noise level goes significantly down. This is shown by the image on the right. The cut-section, like before, shows intensity profile. The flat-topped region is significantly higher compared to nearby regions. This happens as the noise level is reduced leading to a higher SNR. The resultant image is ‘cleaner’ and has high visibility.

Figure 3: The basic premise behind stacking is to create a superposed image from several sub-exposures. 

Figure 4: Noise profile and star signal (Left) In a single sub-exposure (Right) In stacked image

Several Mac or Windows based softwares are available for stacking. For Windows OS, Sequator and DeepSkyStacker are quite popular. For Mac OS, a paid utility is the StarryLandscapeStacker. For each of these softwares, several tutorials are available on the net, and so we will not go into that here. We hope to highlight differences in these softwares in later essays.

Figure 5: Stacking in Sequator.

नक्षत्राणां ज्योति: मन: प्रचोदयात्

May the light of the stars illuminate your mind

 Notes:

1. AK dedicates this series of posts to Annapoorna.

About the authors: (First) Aloke Kumar is currently an Associate Professor at Indian Institute of Science, Bangalore. He tweets at @aalokelab

(Second) Shubhanshu Shukla is an amateur astrophotographer.

*'I' in an article refers to the first author of that article

Opinion/views expressed are purely personal and do not reflect the opinion/views of employers.