Skip to content

Explore faster implementations of median filter #7475

Open
@lagru

Description

Description:

Opportunities for improving the performance of specific algorithms was mentioned several times in the faster scientific Python discussion. Especially with regards to median filtering there are more specialized algorithms as described in this blog post on Median filtering by @crisluengo.

Not saying that this is our highest priority, but in case someone feels "the itch" in particular, idea and resources are recorded here. :D

Activity

naik24

naik24 commented on Aug 6, 2024

@naik24

Hi, can I work on this issue? I will refer to the blog post and implement the different techniques and submit a pull request accordingly.

lagru

lagru commented on Aug 12, 2024

@lagru
MemberAuthor

Hey @naik24, welcome. Thanks for offering to work on this. 😊 Do you have experience with this kind of work?

Just to be clear up front about this. This looks like a big endeavor and we currently have limited resources on scikit-image's side to mentor you. So this would involve a bit of independent investigation and work on your side. But of course we would be willing to help you along with high-level feedback and advice!

For scikit-image we would probably go for the Binary tree based algorithm which would have to be implemented in terms of scikit-image (see the DIPlib implementation for a reference on complexity). Of course, you are free to explore other even better solutions. Though, we would prefer an algorithm that can completely replace our current one in terms of features and performance.

If that sounds alright to you, go for it! :D

Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Metadata

Assignees

No one assigned

    Type

    No type

    Projects

    No projects

    Milestone

    No milestone

    Relationships

    None yet

    Development

    No branches or pull requests

    Issue actions

      Explore faster implementations of median filter Β· Issue #7475 Β· scikit-image/scikit-image