FOAK: Any ImageMagick expertise in the house?

Domènec wrote:
> "Higgins" <the.best.names.are.gone DeleteThis @gmail.com> escribió en el mensaje
> news:13s2t1ef3p0g794@corp.supernews.com...
>
>> I'm trying to figure out a way to compare two fairly simple images and
>> generate a metric which reflects the similarity of the shapes within those
>> images. The images are monochrome, so colour isn't relevant.
>
> Need more than [1]?
>
> [1] http://www.imagemagick.org/script/compare.php
>
>
>
Been there and I can't quite seem to get what I need but this is new
territory for me. It seems that compare looks at pixels and colours,
rather than edges, so most of the things I am trying to compare result
in fairly similar metrics.

What I'm trying to do is gain a level of confidence that the shape
inside the image is the same (think OCR but more limited).
 >> Stay informed about: FOAK: Any ImageMagick expertise in the house? 

"Higgins" <the.best.names.are.gone.RemoveThis@gmail.com> escribió en el mensaje
> Domènec wrote:

>>> I'm trying to figure out a way to compare two fairly simple images and
>>> generate a metric which reflects the similarity of the shapes within
>>> those images. The images are monochrome, so colour isn't relevant.
>> Need more than [1]?
>> [1] http://www.imagemagick.org/script/compare.php
> Been there and I can't quite seem to get what I need but this is new
> territory for me. It seems that compare looks at pixels and colours,
> rather than edges, so most of the things I am trying to compare result in
> fairly similar metrics.
> What I'm trying to do is gain a level of confidence that the shape inside
> the image is the same (think OCR but more limited).

Me understands. Looks like ImageMagick compare is not the tool you need, all
I can say.
 >> Stay informed about: FOAK: Any ImageMagick expertise in the house? 

Higgins wrote:

> I'm trying to figure out a way to compare two fairly simple images
> and generate a metric which reflects the similarity of the shapes
> within those images. The images are monochrome, so colour isn't
> relevant.
>
> Any ideas?


Never used IM to do anything other than convert from one format to
another, but if I were doing this I would:

Easy: If they're filled in shapes, xor the pictures, count the
resulting set pixels.

Medium: If they're straight line drawings, measure the length angle,
and intersection of the lines.

Hard: If they're line & curve drawings, add curve radius & intersection
to the above.

Very hard: build a set of weak, democratic, neural nets trained to
measure various kinds of similarity.
 >> Stay informed about: FOAK: Any ImageMagick expertise in the house? 

Simian wrote:
> Higgins wrote:
>
>> I'm trying to figure out a way to compare two fairly simple images
>> and generate a metric which reflects the similarity of the shapes
>> within those images. The images are monochrome, so colour isn't
>> relevant.
>>
>> Any ideas?
>
>
> Never used IM to do anything other than convert from one format to
> another, but if I were doing this I would:
>
> Easy: If they're filled in shapes, xor the pictures, count the
> resulting set pixels.
>
> Medium: If they're straight line drawings, measure the length angle,
> and intersection of the lines.
>
> Hard: If they're line & curve drawings, add curve radius & intersection
> to the above.
>
> Very hard: build a set of weak, democratic, neural nets trained to
> measure various kinds of similarity.
>
I suspect that, if I need to go beyond the "easy" option, I'll be going
for "give up" .
 >> Stay informed about: FOAK: Any ImageMagick expertise in the house?