On Jun 18, 2004, at 6:07 PM, Jasper Phillips wrote:
A huge one, probably. I'm sure it would also entail quite a bit of bloat, especially on Win32 and OS X where you can't assume GTK and whatnot is already there.On Fri, 18 Jun 2004, Bob Ippolito wrote:Not to hard, provided I could get pygimp to work. I'd have to compile andOn Jun 18, 2004, at 5:34 PM, Jasper Phillips wrote:It is however very frustrating that I have to reinvent the wheel! It
really puts into perspective the value of all the amazing LGPL code I'm
using, without which I'd be stuck.
Yeah, but the functionality that you want from GIMP is probably really hard to decouple anyway. You want a wheel but you have an 18 wheeler :)
ship it to, which would be a headache.
Regardless of optimization, I don't think PIL has the functionality you would need. Clothify isn't exactly a standard filter, and I doubt it's as easy as a convolution (if PIL even gives you that).Hmmm, are you implying that PIL isn't as optimized as these things? I'veIt can't be that hard to look at the code they're using and write a similar algorithm on top of Numeric/numarray/scipy/etc.
been wondering about that, for the amount of code that PIL has for filtering
is negligible, and at a cursory glance it sure appears unoptimized...
Normally I don't give a damn about execution speed, but image filtering
seems like an obvious bottleneck judging from how long it takes GIMP to
clothify.
Attachment:
smime.p7s
Description: application/pkcs7-signature