Different value of Haralick Feature and The Direction

[Avicienna Ulhaq]

Hi there !

i'm trying to measure haralick feature and compare it with imageJ plugin "Texture analysis" in here.

it give different result. as you can see in the picture below is result of JFeatureLib 

and this one is result of ImageJ "Texture Analysis"

one of my assumption is both of them using different direction to calculate haralick feature. but i cant found direction properties in JFeatureLib.

Thanks for your help ! :)

[Franz Graf]

Heyo,

sorry to answer that late. I was on some business + holiday trip.
Hm yea this looks strange.
I rmemeber that we took the Haralick-computation from legacy code that
was already around at the department. The direction (at least according
to the docs) should be the first param and thus be found in features[0]
(Haralick.java L:232).

A little difference in the computation of the coocurrence matrix
(Haralick.java L:472ff) could also lead to different results. Surely
also the value for HARALICK_DIST and how a grayscale scaling is
performed (L:456) :-/

Maybe this helps a bit further.

Good luck!
Franz

Am 12.05.2015 um 02:42 schrieb Avicienna Ulhaq:
> Hi there !
>
> i'm trying to measure haralick feature and compare it with imageJ plugin
> "Texture analysis" in here

> <http://rsb.info.nih.gov/ij/plugins/texture.html>.

>
> it give different result. as you can see in the picture below is result
> of JFeatureLib
>

> <https://lh3.googleusercontent.com/-6kVGNYHxwPI/VVFLyZ8UV4I/AAAAAAAAAt0/MXo3aosqGnU/s1600/JFeatureLib_haralick.png>

>
> and this one is result of ImageJ "Texture Analysis"
>

> <https://lh3.googleusercontent.com/-J0Pv_WNLY-U/VVFMJd-xNfI/AAAAAAAAAt8/JtE1223Eb0s/s1600/ImageJ_Texture.png>

>
> one of my assumption is both of them using different direction to
> calculate haralick feature. but i cant found direction properties in
> JFeatureLib.
>
>
> Thanks for your help ! :)
>

> --
> You received this message because you are subscribed to the Google
> Groups "JFeatureLib" group.
> To unsubscribe from this group and stop receiving emails from it, send
> an email to jfeaturelib...@googlegroups.com
> <mailto:jfeaturelib...@googlegroups.com>.
> To post to this group, send email to jfeat...@googlegroups.com
> <mailto:jfeat...@googlegroups.com>.
> For more options, visit https://groups.google.com/d/optout.

--
Dr. Franz Graf
http://www.Locked.de