The Best Way to Reduce Noise at Night
Updated: Feb 21, 2021

This is a stitched, multi-row panorama of the Milky Way over a sandstone tower in Chesler Park, Canyonlands National Park, Utah. I shot one row for the land and two rows for the sky. I shot four frames at each camera position. I used Stack Mode>Median to reduce noise in each group of land images. I used RegiStar software to re-align the stars in each group of sky images, then used RegiStar's equivalent of Stack Mode>Median to reduce noise in the sky images. I used Lightroom to stitch together all the completed sky images. Then, as a separate operation, I stitched all the completed land images together. Finally I blended the sky panorama and land panorama together in Photoshop.
Noise is the great enemy of night photographers. Sure, the noise-reduction sliders in the Detail panel in Lightroom work pretty well, but the result is always a compromise. Noise-reduction software attempts to suppress the high-frequency variation in brightness and color that is unwanted noise while preserving the high-frequency variation in brightness and color that is desirable texture and detail. Inevitably, the more you reduce noise, the more you lose fine detail.
Instead of attempting to smooth out noise, delete it using Photoshop’s Stack Mode>Median. The basic idea is to shoot a series of identical frames and stack them as layers in a Photoshop file. Photoshop can then drill down through the layer stack, comparing the value of each pixel to the value of all the other pixels directly above and below it. Photoshop then chooses the median value—the number in the middle of the range—and displays it.
Here’s a simplified example. Let’s say you shoot four identical images. As Photoshop drills down through the layer stack at a particular point in the scene, it finds four pixels. Three have RGB values averaging out to 100. The fourth has RGB values averaging out to 200. The pixels with an average RGB value of 100 are pixels showing the correct brightness and color. The pixel with an average RGB value of 200 is noise—a random and undesirable variation from the true value you want to capture. The median value of that set of four numbers is 100, so that is the value Photoshop displays. Instead of smoothing out the noise, you’ve removed it.
In the real world, of course, the effect isn’t quite that dramatic. In my testing, I’ve found that shooting four frames at ISO 6400, then reducing noise with this method, produced an image that looks like it was shot at ISO 1600. Shooting 10 identical frames shot at ISO 6400 then stacking them to reduce noise produced an image that looked like it was shot at ISO 800 or even lower. Results from your camera may be different, of course.
It’s easiest to apply this technique to the land, rather than the sky, since the land doesn’t move between frames. Fortunately, the land portion of a night image is usually the part that shows the most noise. Noise is always worse in the deep shadows, and the land will always be darker than the sky. In addition, noise-reduction software does a better job with the sky than it does with the land. Usually you can find a setting that will smooth the sky while also preserving faint stars. Applying heavy noise reduction to the land, on the other hand, can easily smooth out textures to the point that the land looks like injection-molded plastic.
You can also shoot multiple frames using the correct exposure for the sky, but the stars will move enough between frames that they won’t align when you stack the images in a layered Photoshop file. Photoshop’s Auto-Align Layers utility is designed to identify features in the land portion of the frame and make them align. It won’t work to align stars. Deleting the land from all the frames won’t help; in my experience, Photoshop will simply report that it can’t align the images at all. Registering the stars requires specialized astronomical software. As I write this, the best programs I’ve found for this purpose are RegiStar from Auriga Software (PC only) or Starry Landscape Stacker (Mac only). There may be other programs available by the time you read this. Once the stars are aligned, you can use the Stack Mode>Median-Mean utility built into RegiStar or the noise-reduction utility built in to Starry Landscape Stacker to reduce noise. You’ll then need to stack the good-sky and good-land images in Photoshop and blend them. The process of aligning the stars turns the land into a blurry mess, so you can count on lots of cloning along the skyline to create a clean boundary between the two images.
Here, in detail, is the procedure for using Stack Mode>Median to reduce noise in the land. Start by shooting at least four, and preferably more, identical frames using the correct land exposure. The more frames you shoot, the more you’ll be able to reduce the noise. Once you’ve collected the necessary frames, download them to Lightroom, select all of them, and open one in the Develop module. Be sure Auto-Sync is enabled so that a change applied to one will be applied to all, then edit the good-land images to taste. Don’t apply any luminance noise reduction at this stage. Experiment with turning off color noise reduction as well, then applying it again once you’ve created the finished image. You may find this helps preserve additional detail. With all the images still selected, right-click on one of them and choose Edit In>Open as Layers in Photoshop. In Photoshop, select all the layers by clicking on the top layer and Shift-clicking on the bottom layer. Right-click on one of the layers and choose Convert to Smart Object. Then choose Layer>Smart Objects>Stack Mode>Median. To reduce file size, right-click on the Smart Object layer and choose Rasterize Layer. That’s it—you’ve just learned the best way to reduce noise in night landscapes.

This image shows a small portion of a single frame, enlarged to 100 percent. I shot it at 2 minutes, f/2.8, ISO 6400 with a Canon 5D Mark IV and did not apply any noise reduction in Lightroom. Note the high level of noise visible throughout the image.

To create this image, I shot four identical frames at two minutes, f/2.8, ISO 6400 with a Canon 5D Mark IV, then stacked the images in Photoshop and used the Stack Mode>Median method described above to reduce noise while preserving the desirable texture in the rock. I did not apply any additional noise reduction in Lightroom.
Want to know when new blog posts are released? Please join my mailing list!
Want to learn more? Get information on the private and group landscape photography workshops I teach.





I found this article informative and practical for homeowners. An experienced Interior Remodeling Contractor can transform living spaces through thoughtful design, efficient planning, and quality workmanship. Professional guidance often helps achieve attractive results while minimizing disruptions during construction projects.
I found this discussion very useful for understanding online visibility and engagement. Tools that Discover Local Businesses make it easier for customers to explore nearby options, compare services, and support community growth. Accessible information benefits businesses and consumers alike today.
This is a really helpful breakdown of a complex astrophotography workflow, especially the way the author explains using median stacking and RegiStar to handle noise in both land and sky separately. The step-by-step approach to stitching and blending makes it feel achievable rather than overwhelming. If you enjoy discussing techniques like this, image chat also has related content worth checking out.
Great explanation of using Stack Mode > Median to remove noise while keeping detail! If you ever translate these complex masking skills into video projects or night timelapses, I highly recommend ExportAlpha. It's an AI tool that makes it incredibly easy to export alpha channels (like ProRes 4444 MOV or transparent VP9 WebM) for seamless compositing. Huge time-saver!
This is one of the clearest explanations of Stack Mode>Median I've come across, and the RGB example really clicks. The idea that Photoshop drills down through the layer stack, finds four pixels with values of 100, 100, 100, and 200, and simply discards that outlier of 200 because the median is 100 — that's a much more intuitive way to understand it than "averaging." You're not blurring anything; you're surgically removing the random variation while keeping the genuine detail intact. That distinction between smoothing noise and deleting it is exactly what most noise-reduction tutorials gloss over.
It strikes me that this whole article is really about planning before you shoot rather than fixing things afterward — deciding your ISO, frame…