Introduction

Summertime landscape astrophotography in the southwest can be seriously sweat-inducing. Chances are that if you’re standing in one place and still feel hot, so does your camera sensor!

The aptly named ‘hot pixel’ is a defective pixel or grouping of subpixels that becomes most noticeable when the temperature of the sensor gets too hot. The sensor temperature is influenced by operation duration and the ambient air temperature, among other factors. The high temperature makes the sensor more prone to current leaks, which disproportionately affect the values of these defective pixels.

There’s nothing wrong with your camera! It is completely normal for cameras to behave like this, even for high-end/prosumer deep space cameras.

Traditionally, deep-space setups make use of a cooler and dark frames to get rid of hot pixels. Since hot pixels don’t change places between images, deep space astrophotographers can easily remove them by stacking a series of ‘dark frames’. For us landscape-astrophotographers however, dark frames are rarely a realistic option.

Here’s why: practical applications of photography like timelapses or image stacking (both covered in this example) take a lot of time to complete. During this time, the ambient temperature can fluctuate quite a bit, especially for something like a timelapse that is set to run all night. As a result, the single-image distribution of hot pixels can vary quite a bit. Should you take dark frames in these conditions, you will find that your dark stack is unrepresentative of the conditions when you were taking your photos. In other words, since our relatively convenient DSLR and mirrorless cameras don’t come with built in temperature regulators like the astrocams do, we can end up subtracting too many or too few hot pixels using our post-hoc dark frames, leaving us back at step 1 with several visible artifacts– some appearing quite large.

This is where synthetic darks come in hand.

The advantages of using synthetic darks:

• Image local – no need to worry about variance
• Easy to create on many different image editing software
• Practical for timelapse cleaning
• Practical for transforming a timelapse (motion or static) sequence into a stacked image
• They preserve surrounding noise without introducing any additional noise
•  Can still be incorporated into a non-destructive workflow
• Can be applied to thee image globally – no need to mask with brushes.
• Expandable to cover several different editing situations

Tutorial

Disclaimer, techniques making use of synthetic darks alone will work best on foreground images. To extend this technique to the sky portion of the image, you’d have to remove the stars using an Ai tool like StarNet (free) or StarXterminator (paid), or use a process called Cosmetic Correction that can be found in SiriL (free) and Pixinsight (paid), which uses statistical analysis of a dark frame or dark stack to remove hot pixels anywhere in the image. I would highly recommend SiriL and StarNet to those starting out as both are incredibly powerful tools. You can accomplish high level edits using these without ever having to move to the paid software, I say as someone who owns both.

The example I will be using to demonstrate synthetic darks and noise reduction is a 10 image sequence from a larger motion timelapse I took of the aurora that I have turned into a stand-alone image. This requires image alignment which I will be skipping. The movement of elements in the scene also increases the effectiveness of the stacking for noise reduction. For a static timelapse stack, I would instead opt to use the synthetic dark as a mask to content-aware fill the hot pixels. For static timelapse video cleaning, a stacked synthetic dark could be used to remove the average hot pixel variation from the full video sequence as a subtraction overlay. Since a synthetic dark stack is essentially a model for the video’s hot pixels, you shouldn’t see much improvements to the stack after incorporating ~30 images.

The lowered noise in the comparison image at the beginning of the post is purely a result of image stacking, which is inherently Ai-free.

Step 1

Using photoshop, go to File -> Scripts -> Load images into stack, and select as many images as you would like. I will be selecting 10 to try to limit the scratch disk space used. These can be raw images or tiffs. I wouldn’t recommend much beyond simple color correction as far as edits go. Before exporting tiffs or closing camera raw, I would recommend sliding distortion correction to zero (0) to avoid ringing artifacts when stacking.

For images taken from a motion timelapse sequence, check ‘attempt to automatically align source images’ before clicking ‘OK’. Images shot on a static tripod may be able to avoid this step.

Below is a simulated stack to illustrate the problem, and where the image-local-ness of synthetic darks comes into play.