Denoising & segmentation¶
These commands clean up and segment MR images. Two of them remove noise from an image with a total-variation prior, which smooths regions while keeping their edges; the third segments brain tissue with a model that also corrects the bias field of the scanner and aligns the image to a template.
Several contrasts of the same subject can be denoised together so that they share their edges. The default template of the segmentation is a brain atlas, so it expects a head image; it can combine several contrasts.
| Command | Summary |
|---|---|
denoise |
Total-variation denoising of one or several images, with the noise level and the regularisation under your control. |
denoise_mri |
Joint total-variation denoising of co-registered MR sequences, with noise level and regularisation estimated from the images. |
uniseg |
Unified Segmentation, as in SPM12: tissue classes, bias field and template alignment, estimated together. |
Choosing between the two denoisers¶
Use denoise when you want to choose the noise level and the
amount of smoothing yourself, or when your images are not MR volumes. Use
denoise_mri when you have MR sequences and prefer
parameters derived from the image intensities. The page of denoise_mri
compares the two in detail.
See also¶
- Registration: bring several contrasts into the same space before denoising them jointly.
- Images: smoothing, inpainting and other image operations.