Detection Binning¶
Star detection is calibrated for stars whose in-focus HFR is roughly 2 to 4 pixels. Every pixel-based setting in the detector assumes that: the minimum bounding box, the minimum HFR, the noise-reduction radius, the number of structure layers, the background box expansion, the centering tolerance. At long focal lengths an in-focus star is far bigger than that, so those defaults are out of range, detection has many more pixels to chew through, and each pixel carries less signal than it needs to.
Detection Binning resamples the frame before detection so star sizes land back in that range. It applies to detection only. The image NINA displays, and every HFR, FWHM and star position Hocus Focus reports, stay at the resolution the camera captured.
The setting¶
Find it in the Star Detector tab, below the other detection settings. It applies in both Simple and Advanced mode.
Software binning applied to the frame for star detection only. Star detection is calibrated for in-focus stars of roughly 2 to 4 pixels; at long focal lengths stars are much larger than that, which puts the pixel-based settings out of range. Binning the frame for detection brings star size back into range and improves signal to noise per pixel. The image you see and the HFR values reported stay at the resolution the camera captured. This is not NINA's Auto Focus Binning, which changes how the camera captures the auto-focus frames and should match the binning you image at; the two multiply
Default: 1x1 (Off) • Range: 1x1 (Off), 2x2, 3x3, 4x4.
Nothing chooses the factor for you. A line appears under the dropdown only when there is something to act on: no measurement yet, a factor change, or a measurement below the range the detector is calibrated for:
Run an auto-focus to get a recommendation
Measured in-focus HFR 6.1 px - 2x2 recommended
Measured in-focus HFR 1.4 px - below the 2-4 px range
Once your setting matches the factor the measurement calls for, and the measurement is not below the calibrated 2 to 4 px range, the line disappears: there is nothing to act on. Hover it for the reasoning, including when the measurement was taken. In the Optimization Wizard the same line sits to the right of its dropdown, under the same rule, and its summary offers to re-run the search at the recommended factor.
The recommendation is advisory; nothing enforces it.
Why there is no Auto
An automatic factor would change how frames are analyzed the moment the plugin updated, and every pixel-based setting you had already tuned would silently be measured against different pixels. The setting is explicit for the same reason the wizard makes you re-optimize after changing it: the factor and the settings tuned at it belong together.
Where the recommendation comes from¶
It is measured, not estimated. The number is the final HFR of your last auto-focus run: a real exposure taken at the focuser position the run settled on. Accepting a live sweep in the Optimization Wizard updates it too, from the sweep's fitted curve. The factor is then whichever brings that measurement closest to 3 px:
| Measured in-focus HFR | Recommended |
|---|---|
| below 4.5 px | 1x1 (off) |
| 4.5 to 7.5 px | 2x2 |
| 7.5 to 10.5 px | 3x3 |
| 10.5 px and above | 4x4 |
Pixel scale cannot answer this
The recommendation is never derived from pixel scale and an assumed seeing figure. Plausible seeing spans roughly 1.5″ to 4″, a factor of 2.7, which is wider than the whole 1x1-versus-2x2 decision margin: on a 0.28″/px rig the answer flips at about 2.3″ of seeing. An assumed figure would decide the recommendation, not your rig.
Changing the rig discards the measurement¶
The measurement is a star size in captured pixels, so it only means anything at the pixel scale it was taken at. A 5.8 px star at 2800 mm is 2.9 px at 1400 mm — the same sky, a different recommendation. So Hocus Focus watches the active profile's focal length, pixel size and Auto Focus Binning, and discards the measurement whenever they move the pixel scale. The line under the dropdown goes back to asking for an auto-focus, and the next auto-focus refills it.
It compares the resulting pixel scale rather than the three values, so a change that cancels out — doubling focal length and doubling capture binning — keeps the measurement. Edits arrive one at a time, though, so making those two changes one after the other still discards it on the first: nothing can know a compensating second edit is coming. The cost is one auto-focus, which is the right side to err on.
Switching profiles does not discard anything. Each profile stores its own measurement, so a switch simply moves to that profile's own value. Per-filter Auto Focus Binning overrides are not watched — only the global setting is — so re-run an auto-focus by hand after changing one of those.
This is not NINA's Auto Focus Binning¶
NINA has its own Auto Focus Binning setting, in Options → Focuser (and per filter in the filter wheel settings). The two are different things and they multiply.
| NINA Auto Focus Binning | Hocus Focus Detection Binning | |
|---|---|---|
| What it changes | How the camera captures the auto-focus frames | How the frame is analyzed, after capture |
| Set it to | The binning you image at, so focus is found for the frames you actually shoot | Whatever brings star size into the detector's range |
| Effect on the displayed frame | Smaller frame, larger pixels | None |
| Effect on reported HFR | Reported in the captured (binned) pixels | None; still reported in captured pixels |
Set Auto Focus Binning to match your imaging binning, and set Detection Binning from the recommendation. If you raise Detection Binning while Auto Focus Binning is above 1x1, Hocus Focus explains the difference and offers to set the NINA setting back to 1x1, listing the global value and any per-filter overrides it would clear. Answering No keeps both, and the two factors stack.
The recommendation accounts for camera binning automatically, because the HFR it reads was measured on frames the camera had already binned. On one simulated 5600 mm rig it asks for 3x3 unbinned and 2x2 with the camera at 2x2.
Changing it in the Optimization Wizard¶
The Optimization Wizard shows Detection binning next to Capture binning on its confirmation panel, and it is settable there: the whole search is tuned at whatever factor is in effect when the run starts, so that is the last useful moment to choose it. It is the same setting as the one on the Star Detector tab, edited in place.
After a run, the summary reports the factor your measured focus curve calls for. If it differs, the only action offered is Optimize again at NxN, which repeats the search on the frames already captured at the new factor. Nothing is saved until you Accept that result; the factor and the settings tuned at it are then applied together. The factor is never applied on its own, because settings tuned at one factor are not valid at another.
What is and is not rescaled¶
Detection runs on binned pixels; everything it returns is converted back to captured pixels before it leaves the detector. So HFR, star positions, bounding boxes, PSF sigma and the rejected-candidate boxes drawn in the annotation overlay are all in the frame's own coordinates, and the auto-focus curve means the same thing at any factor.
Two consequences:
- Star centers are coarser. A star's position in captured pixels is quantized to roughly the binning factor. That is the trade for the signal-to-noise gain. Auto-focus depends on HFR, not on sub-pixel centroids, so this does not affect focus accuracy.
- HFR is not identical between factors. Binning raises the signal-to-noise per pixel, so more of each star's outer flux clears the measurement threshold and the measured half-flux radius creeps up: on synthetic frames, about 5% at 2x and 21% at 3x. That is a scale factor across the whole focus curve, so it does not move best focus, but HFR numbers from different factors are not directly comparable. Changing the factor is a good moment to re-run the Optimization Wizard.
Detection also crops the frame to a whole multiple of the factor first, so up to three rows and columns at the right and bottom edge are dropped. Stars there would be rejected by the border gate anyway.
When to change it¶
Follow the recommendation unless you have a reason not to, and re-run an auto-focus after changing optics so it stays current. Reasons to override:
- Stay at 1x1 if you want detection to see every pixel — chasing a specific detection problem, or comparing against an older run's HFR values.
- Go higher than recommended if you routinely image in worse seeing than the night the measurement was taken, since the measurement is one night's answer rather than a long-run average.
Whichever you pick, re-run the Optimization Wizard afterwards. Every pixel-based setting is measured in binned pixels, so a factor change makes the tuned values mean something different.
Feedback: in the Star Detection Results panel, a factor that suits your rig raises Total Detected (fainter stars clear the threshold) and cuts detection time, without a jump in the Too Small or Too Distorted counts.
Where it sits in the pipeline¶
Detection binning is an EARLY parameter: it resamples the frame before candidate finding, so changing it forces a full re-detect and the Optimization Wizard cannot reuse a cached context across factors. Hot-pixel filtering deliberately runs before the resample, at full resolution, because a hot pixel averaged into its block is no longer the isolated outlier the filter looks for.