Re: [boost] GIL - Generic Image Library Review - Begins October 5

From: Lubomir Bourdev (lbourdev_at_[hidden])
Date: 2006-10-06 12:57:10

Andy Little wrote:
>The domain that the library can be used in is very narrow. To be seriously used
>for image recognition as was previously suggested an application, my guess is
>that the library would need the ability to apply arbitrary transforms, including
>other than 90 degree rotations, and  interpolation of points, stereoscopic
>vision etc.

That is a fair point - having image processing algorithms would make GIL more useful.
 
And we would like to have image processing algorithms at some point in the future. But writing these in a generic and efficient way is a huge undertaking. We are hoping the open source community will join us in writing a future numeric extension to GIL (and, if GIL makes it into boost, a future boost GIL-algorithm extension proposal)
 
That said, we don't want the lack of image processing algorithms to be an impediment for those who want to use them. This is why we recently provided a small numeric GIL extension that gives you a starting point to writing image processing algorithms. You can get it off GIL's main download page (step 5):
 
http://opensource.adobe.com/gil/download.html
 
The algorithms there are not well documented and not optimized for performance. But you can do a convolution and generic resampling (nearest-neighbor and bilinear interpolation). That lets you do things like blurring, sharpening, rescaling the images, arbitrary-degree rotation, etc. We have a sample file that shows you how to do this.
 
Even if we had a fully-optimized and comprehensive image processing extension to GIL, it is probably not a good idea to include it in this boost proposal. If you think GIL is big now, imagine how much bigger it will be with the algorithms.
 
Lubomir