Hi Ru,
your image seems to be a grayscale image with a color lookup
table, so you should use the original grayscale data.
Also, I have the impression that there are grain boundaries in
it with very different contrast (are the weak ones subsurface?)
and it is almost impossible to decide when looking at it what
is strong enough to be taken as bona fide grain boundary.
Usually, if the human eye+brains (with millions-of-years evolution
of image processing) cannot decide, an automated algorithm
cannot do any better.
If your original data are not too noisy (on picasa, your images
are badly degraded by jpeg artifacts), you can try playing
with Unsharp Mask to enhance the grain boundaries before
further processing.
A more general remark - I am not deep enough in image
processing for this:
I wonder whether there is any good filter that enhances linear
features, but not pointlike ones, irrespective of the orientation
of the linear features. E.g., something that is like running
many Gabor filters with different orientations and combining
them in a nonlinear fashion.
Michael
________________________________________________________________
On 13 Jun 2007, at 06:09, Ru Wang wrote:
> Dear All,
>
> I have no idea to find the grain edges on a color image.
>
http://picasaweb.google.com/wangru.ustc/Image_ProcessAlbum> If you have any suggestions, I will be very appreciated.
>
> Thanks.
>
> Ru
Locked
