Edge preserving filters... Which filter(s)?

I have been using ImageJ/Fiji on nano-particle images in an attempt to characterize size, shape and obtain distributions. To this end I have posted here previously, watched some of the training video’s and read their associated presentations. I am curious to know if the filters which are known to preserve edge information can be rank ordered or delineated in this manner. My searches for this information in this forum did not find it presented in such a convenient manner. I also turned to Google Scholar and found this to be a very active area of research.

Off the top of my head I would suggest the bilateral filter: https://imagejdocu.tudor.lu/wishlist/completed/bilateral_filtering

Also the median filter is edge preserving: https://imagej.nih.gov/ij/docs/guide/146-29.html

As a more advanced filter the non-local means has also good properties: https://imagej.net/Non_Local_Means_Denoise
but it is a bit computationally expensive.

5 Likes

@schmiedc - Thank you for the response. I knew the median filter was in this category. The other two you mention are new to me. I think the Kawuhara filter is in this category as well and I located this today

And two more:

– Michael

2 Likes

Some more information regarding edge-preserving smoothing can be found in:

  • Chapter 5 (Edge-preserving Smoothing Filters) in Burger&Burge, Principles of Digital Image Processing - Advanced Methods.
    All the filters described there are implemented in ImageJ and are available as jar files.

  • Also see Hybrid Median Filter in ImageJ plugins page.

Peter

1 Like

I am not an expert of the filters, however, I usually have to filter the slices from neutron tomography scans. Sometimes it is a really demanding task to preserve the edges with reasonable smoothing elsewhere, and I use a program developed one of my colleagues. This code applies the so called ISS (inverse scale space) filter and the Diffusion filter. I like the ISS because it helped a lot in my case. I have not found such filter developed in imagej but perhaps there is something.
Here you can find some details of the filters: https://github.com/neutronimaging/imagingsuite/wiki/KipTool-manual:-AdvancedFilterModules

Zoltan

2 Likes

… and not to forget … the Mean Shift

https://imagej.nih.gov/ij/plugins/mean-shift.html

Hi again,
I have seen a paper in Nature introducing a FIJI plugin, which is useful for denoising electron microscopy data. I can imagine that it could be used for other type of images. I tried it just for fun with my images, and it works. Here is a link to the website:
https://bioimagingcore.be/DenoisEM/index.html

I hope it helps a little bit.

Zoli