Log file Viewer

[15:00:43] <lasconic> Hi

[15:01:34] <lasconic> I have these sets of images

[15:01:37] <lasconic> http://vtest.musescore.org/187236d8/vtest.html

[15:01:58] <lasconic> I would like to compare first and second columns better than what we are doing right now

[15:02:09] <lasconic> (third row)

[15:02:25] <lasconic> can anyone give me some hints ?

[15:36:16] <wiredfool> what would constitute a better comparison?

[15:38:33] <lasconic> wiredfool: the main difference is that one image is bolder than the other one

[15:38:45] <lasconic> if it's the only thing, then the images are identical

[15:38:50] <wiredfool> ah

[15:39:46] <lasconic> wiredfool: does it make sense?

[15:39:47] <wiredfool> are you worried about 1px vertical alignment issues?

[15:39:59] <lasconic> not really

[15:41:23] <lasconic> wiredfool: what I want to detect is this

[15:41:23] <lasconic> http://vtest.musescore.org/8088634e/vtest.html#bravura-2

[15:42:04] <lasconic> or this http://vtest.musescore.org/8088634e/vtest.html#accidental-5

[15:42:23] <wiredfool> are all of these supposed to come up equal?

[15:43:15] <lasconic> when everything works fine yes

[15:43:26] <lasconic> the two I linked should raise a red flag

[15:43:31] <lasconic> so not identical

[15:43:42] <lasconic> all the others are ok

[15:44:40] <wiredfool> I think you're getting into computer vision, rather than image manipulation

[15:44:59] <wiredfool> I think that for some of it, an image erosion step after the subtraction might work

[15:45:14] <wiredfool> where you erode anything that's less than n pixels wide

[15:46:14] <wiredfool> scikit.image may help

[15:51:13] <lasconic> image erosion

[15:51:15] <lasconic> ok

[15:51:20] <lasconic> wiredfool: I will take a look

[15:51:28] <lasconic> maybe image magick can do that for us

[15:52:47] <wiredfool> Dilation would be another one to look at

[15:52:52] <wiredfool> they

[15:53:09] <wiredfool> 're roughly inverse operations

[15:53:15] <lasconic> I see

[15:53:18] <wiredfool> http://en.wikipedia.org/wiki/Erosion_(morphology)

[15:53:38] <wiredfool> I think it's in scikit, or possibly numpy

[15:54:27] <wiredfool> there's more complicated stuff ehere, with pretty pictures: http://scikit-image.org/docs/dev/auto_examples/

[15:54:50] <wiredfool> http://scikit-image.org/docs/dev/api/skimage.morphology.html?highlight=erosion#skimage.morphology.binary_erosion

[15:57:34] <lasconic> wiredfool: thank you I will take a look