Raven filter#

Description

Raven filter is a part of the module httomolibgpu.prep.stripe for data correction. Raven filter is based on a FFT filter applied in sinogram space to minimise stripes leading to Ring Artifacts in reconstruction. See more about the filter in the original publication [5].

Where and how to use it:

Raven filter can be used in pre-processing, i.e., before the reconstruction to remove the stripes present in the data.

What are the adjustable parameters:

Most important parameters are uvalue and nvalue that define the shape of the filter, as well as, vvalue, the number of rows to be applied to the filter.

Practical example:

In this example we demonstrate how to apply Raven filter to the normalised projection data.

Warning

As with any pre-processing tool, one needs to be careful in applying Raven filter to data. Sub-optimal parameters can potentially lead to other artifacts introduced in the reconstruction and blurring.

../../_images/data_stripes_added_sino.png

Fig. 13 Input (sinogram view) with the stripes added#
../../_images/data_stripes_added_proj.png

Fig. 14 Input (projection view) with the stripes added#
../../_images/raven_filter_sino.png

Fig. 15 After applying Raven filter (sinogram)#
../../_images/raven_filter_proj.png

Fig. 16 After applying Raven filter (projection)#
../../_images/raven_filter_res_sino.png

Fig. 17 Sinogram view of absolute residual between input and output#
../../_images/raven_filter_res_proj.png

Fig. 18 Projection view of absolute residual between input and output#