Clustering of an image by taking into account the spatial context of each pixel (besides its intensity)

Emmanuelle Gouillart emmanuelle.gouillart at nsup.org
Wed Nov 25 07:02:32 EST 2015


Hi Hakim,

I think this example from the gallery does what you want: merging slic
superpixels.
http://scikit-image.org/docs/dev/auto_examples/plot_rag_merge.html#example-plot-rag-merge-py

Best,
Emma

On Wed, Nov 25, 2015 at 03:50:08AM -0800, Hakim Benoudjit wrote:
> Hi Stéfan,

> Thanks for the suggestion.
> I didn't try SLIC algorithm yet, but it seems to me (from the resulting images)
> that it segment images into small regions (superpixels) that could belong
> (visually) to the same object.
> In my case (ideally), these superpixels need to be merged afterwards.
> Do you have an idea on how to achieve the subsequent merging?

> Le mercredi 25 novembre 2015 02:21:00 UTC, stefanv a écrit :

>     On 2015-11-24 01:47:00, Hakim Benoudjit <h.ben... at gmail.com> wrote:
>     > Thanks Juan, I think you're right.

>     > I might have to read the paper on SLIC algorithm to understand how to
>     tune
>     > the "compactness" parameter.

>     You can also use SLIC to label the image, and then compute features of
>     each SLIC region.  Or perhaps that is what is being suggested already, I
>     wasn't sure.

>     Stéfan



More information about the scikit-image mailing list