I knew that this was a thing, but I couldn't figure out how to do it in Photoshop. I found two good references; there are more I'm sure.
- NatureTTL: How to Stack Star Photos to Reduce Noise in Photoshop
- PetaPixel: A Look at Reducing Noise in Photographs Using Median Blending
At some point maybe I'll write an expanded post, but for now I just wanted to make sure I had a record of these references.
 |
| Milky Way over Pigeon Point Lighthouse, October 26, 2019 Stacked frames at ISO 6400 using median blend |
My technique basically followed the work flow in the NatureTTL article, using 5 successive frames. Stars were manually aligned using Photoshop Warp tool, then combined into Smart Object so that Layer > Smart Objects > Stack Mode > Median could be applied. Note that the layer masks really do need to be deleted, not just disabled.
Lighting on the lighthouse is very inconsistent because of various car headlights, and photographers waving their headlights around. A separate group of frames that had "good" light on the lighthouse was median-stacked as well and blended.
In order to recover the grid in the lighthouse windows I used a single frame from earlier in the evening that was not completely blown out by the light.
For comparison, here is my original effort using a single ISO 3200 frame:
 |
| Single frame at ISO 3200 |
Looking at it now, there are still some aspects of the single frame shot that I prefer.
- The drama of the deep shadows
- The multi-frame version now looks somewhat "brown" to me, vs. the pleasant blue of the single frame.
- The lighthouse lighting on the single frame version is pleasantly shaded. The newer one looks flat. In fact the lighting overall seems flat.
But the blended one is much more crisp, and there are an astonishing number of stars. So nothing I can't fix.
Posts
Noise reduction is an essential process in image processing used to remove unwanted variations (noise) while preserving important details. One of the most effective techniques is Median Stack Blending, where multiple images of the same scene are captured and combined to reduce random noise.
ReplyDeleteImage Denoising Projects
In median stack blending, several images are aligned and stacked together. For each pixel position, the median value across all images is selected instead of the average. Since noise is typically random and varies across frames, it gets eliminated during this process, while consistent details remain intact. This technique is widely used in astrophotography, low-light imaging, and scientific imaging, where noise levels are high.
Apart from median stacking, several other noise reduction techniques are commonly used:
Median Filtering: Removes impulse noise (salt-and-pepper noise) by replacing each pixel with the median of its neighborhood.
Gaussian Filtering: Smooths the image by averaging pixel values with a Gaussian kernel, reducing high-frequency noise.
Bilateral Filtering: Preserves edges while reducing noise by considering both spatial and intensity differences.
Non-Local Means (NLM): Reduces noise by averaging similar patches across the image, maintaining texture details.
Wavelet Denoising: Uses frequency decomposition to separate noise from useful signals.
Each technique has its own advantages depending on the type of noise and application. Median stack blending is particularly powerful when multiple images are available, while filtering methods are useful for single-image scenarios.
In summary, noise reduction techniques, especially median stack blending, help produce cleaner, sharper, and more visually accurate images, making them crucial in photography, medical imaging, and computer vision applications.