32 lines
1.3 KiB
Markdown
32 lines
1.3 KiB
Markdown
# Deconvolution
|
||
|
||
Restore an image via regularised deconvolution. Assumes the image was blurred by a Gaussian PSF with the given sigma. Equivalent to Gwyddion's deconvolve.c module.
|
||
|
||
## Inputs
|
||
|
||
| Name | Type | Required | Description |
|
||
|------|------|----------|-------------|
|
||
| field | DATA_FIELD | Yes | Input blurred field |
|
||
|
||
## Outputs
|
||
|
||
| Name | Type | Description |
|
||
|------|------|-------------|
|
||
| restored | DATA_FIELD | Deconvolved (sharpened) field |
|
||
|
||
## Controls
|
||
|
||
| Name | Type | Default | Description |
|
||
|------|------|---------|-------------|
|
||
| method | dropdown | wiener | Deconvolution method: wiener or richardson_lucy |
|
||
| sigma | FLOAT | 2.0 | Gaussian PSF sigma in pixels (0.1–50.0) |
|
||
| regularisation | FLOAT | 0.01 | Regularisation parameter for Wiener filter (1e-6–1.0) |
|
||
| iterations | INT | 10 | Number of iterations (Richardson-Lucy only, 1–200) |
|
||
|
||
## Notes
|
||
|
||
- **Wiener**: Fast, single-pass frequency-domain filter. The regularisation parameter controls the noise/sharpness tradeoff — smaller values sharpen more but amplify noise.
|
||
- **Richardson-Lucy**: Iterative method that preserves positivity. More iterations = sharper result but risk of ringing artifacts.
|
||
- The PSF sigma should match the actual blur in the image. If unknown, start with sigma=1–3 and adjust.
|
||
- For tip-shape deconvolution (non-Gaussian PSF), use Tip Deconvolution instead.
|