reduces noise in image by locally thresholding the discrete cosine transforms of overlapping subimages, using the hard threshold t.


  • FourierDCTFilter works with arbitrary 2D and 3D images.
  • Local DCT thresholding is applied to all overlapping 16×16 partitions of image. The value of each output pixel is the average of the corresponding pixel in all partitions that include the pixel.
  • For multichannel images, color channels are decorrelated by effectively using a 3D DCT.
  • In FourierDCTFilter[image,t], the threshold value t is typically in the range 0 to 1.


open allclose all

Basic Examples  (1)

Reduce noise in an angiography image:

Scope  (3)

Data  (2)

Filter a multichannel image:

Denoise a 3D image:

Parameters  (1)

Specify the hard threshold:

Applications  (3)

Remove salt-and-pepper noise:

Remove Gaussian color noise from an image:

Unsharp masking using FourierDCTFilter:

Possible Issues  (1)

Timings grow rapidly with the dimensions of a 3D image:

Introduced in 2012
Updated in 2014