Dealing with Walking Noise in deep sky images

Walking Noise stems from fixed pattern noise on the camera sensor, which is then exacerbated by imperfect polar alignment (drift) or lack of intentional shifting between exposures (dithering). It appears as faint, coloured "streaks" like "rain" across the background of the final, stacked image, often in the direction of declination drift. The camera's fixed noise moves slightly with each shot. When software stacks the images, these noisy pixels get averaged into lines. Some cameras are more susceptible to walking noise than others.

Dithering is the best preventative solution, which involves intentionally shifting the mount by a few pixels between exposures so noise does not stack on the same spot.

Using dark frames to subtract hot pixels helps mitigate walking noise as does active cooling of the sensor which reduces the thermal hot pixels that create noise. 

Walking noise is often invisible on single frames but appears after stacking and aggressive stretching of the image. 

However, we shall consider here how to deal with a situation where walking noise could not be or has not been avoided. This involves using a software solution to removing the walking noise.

There are a number of de-noising programs; some of them are AI, trained neural networks trained to detect noise and reduce or remove it from an image. They are all able to remove some walking noise but are really intended for the removal of more random noise, so often leave behind traces of the walking noise.

Frankin Marek’s SetiAstro contains a specific walking noise AI de-noiser in his Cosmic Clarity suite. This is trained to detect and remove walking noise specifically and so is likely to make a better job of the de-noising. This software is available for PixInsight and is fully integrated into SetiAstroSuitePro.

Paul Howat, a prominent imaging member of the Swansea Astronomical Society was king enough to provide me with a 16 bit stacked image with walking noise, unguided with a HAC125DX telescope and a QHY585C un-cooled camera. It is a good image but when aggressively stretched, walking noise is revealed.

Click on any image to get a closer view

The stretched image was loaded into SetiAstroSuitePro and Cosmic Clarity Walking noise de-noised using the maximum settings. The image has been zoomed to reveal a portion where the walking noise is clearly visible.

If this image is inspected closely (by clicking on it) the walking noise can be seen.

The de-noised image in SetiAstroSuitePro

If this image is inspected closely (by clicking on it) the walking noise can be seen to have been removed.

Animation blinking between the noisy and de-noised images

Click on the animation to get a closer view.

The Cosmic Clarity Walking noise de-noiser in SetiAstroSuitePro has done a good job of removing the walking noise from the image.

I then processed the de-noised image to show that it is a good image. Of course, image processing is always to the taste of the person doing the processing and the end result may not be what Paul would have produced. Nevertheless, it is an image, free of walking noise and showing the various regions of nebulosity in and between M42, M43 and the Running man nebula.

Processed image

In conclusion, SetiAstro's walking noise de-noiser did a good job of removing the walking noise from this un-guided, un-cooled image.

PlanetarySystemStacker (PSS)

 PlanetarySystemStacker (PSS)

PlanetarySystemStacker is a free, platform independent (runs on Linux, Windows, and macOS), open-source Python based program used in astrophotography to create high-quality, sharp images of planets, the Sun and Moon from sets of image files, AVI or SER files. It analyses and stacks the best frames to reduce noise and reduces distortions. It ranks frames by quality, aligns them globally, and computes a mean image. The workflow includes functionality for analyzing, editing alignment points (allowing manual adjustment), and "blinking" (reviewing) frames to remove poor quality or corrupted frames before final stacking or to compare a processed image with the original unprocessed, stacked image.

PlanetarySystemStacker is available to download via GitHub.

We chose an alternative way of running PSS. The Windows version runs perfectly in WINE in Linux, which made installation and running as easy as it is in Windows.

PSS is another of those programs that have largely escaped my attention and which merits serious consideration for inclusion in one’s aresenal of image stacking and processing software.

PSS can debayer RAW data which means that data need only be 1/3 of the size of RGB data, if they are colour data.

This is good news for Seestar users who capture RAW AVIs of the Sun and Moon. Also, if RAW SER files are captured, the colour files captured can be captured quicker as well as being only 1/3 the size of an RGB image.

Clicking on any image will give an even closer view

We tested the software on Seestar S50 whole disk RAW lunar AVI and on Lunar and H-alpha surface data. Whole disks are best dealt with like planet data.

Screenshot showing the quality curve in relation to the % of frames selected for stacking

Seestar lunar RAW AVI debayered and the best 50% of frames stacked in PSS

Screenshot showing wavelet processing the stacked image in PSS

Final processed and cropped Seestar S50 lunar image

Stacking a surface image

Screenshot showing the lunar stacked surface image

Screenshot showing Wavelet processing the stacked image in PSS

Animation blinking between the original stacked image and the wavelet sharpened image.

The final processed lunar surface image

Processing H-alpha solar data.

The data were captured as two overlapping panels (RAW SERs) covering the whole solar disk using AstroDMx Capture and a Coronado Solarmax II, 60, BF15 H-alpha scope. Each panel SER was processed in PSS and then stitched together in MS Image Composite editor.

Screenshot showing the quality curve in relation to the % of frames selected for stacking

Screenshot showing frequency distribution local warp sizes of alignment points

PSS has gathered all the information it needs to stack the frames. First, at every AP it identifies the sharpest frames to be used for stacking. Since the seeing is a very local phenomenon, frame sets will be different for different APs. Then, for every AP and every contributing frame the local displacement relative to a reference frame is measured, and the shifted AP patch added to the stacking buffer. Clicking OK completes the process.

Screenshot showing Wavelet processing the stacked image in PSS

The stitched two panel stacked images

The image further processed in Gimp3 and colourised

There is not a huge learning curve for PSS, in fact, there is a totally automatic workflow where no user intervention is required. Our results indicated that PSS can do the jobs of software such as Autostakkert! and waveSharp, all in one program. We intend to study this software more closely, and will include PlanetarySystemStacker in our suite of regularly used image processing programs and in our workflows.