ANTs DenoiseImage | Preclinical Webapps

Input image

Drop NIfTI or DICOM files or folder

Denoising

Noise model iUse nonnegative magnitude images. Gaussian, DenoiseImage's default, suits a good signal-to-noise ratio. Signed and phase images are outside the validated scope. Rician suits low-SNR magnitude images, such as diffusion-weighted scans: it removes the upward bias that magnitude noise gives dark tissue.

Advanced settings

Mask optional iDenoiseImage then denoises only the nonzero voxels of the mask; voxels farther than the patch radius from it keep their values. The mask must be on the image grid. A brain mask shortens the run on a whole-head image.

Patch radius voxels iHalf the width of the patches compared around each voxel. DenoiseImage's default is 1 (3 × 3 × 3). Radius 2 compares more of each neighbourhood; on the example it took 3 times as long.

Search radius voxels iHalf the width of the region searched for similar patches. DenoiseImage's default is 2 (5 × 5 × 5). Radius 3 finds more similar patches; on the example it took twice as long.

Output

Drop an image anywhere, or choose an example.

Status Drag and drop files