That is a side effect from the gaussian kernel. If you look at look at a simpler blur kernel, like a box blur, you will see that you take multiple samples of an image, and offset them by a pixel (or more) up/down/left/right and blend together. If one does this, the actual edges of the image end up having a kind of edge because there aren't pixels available "outside the bounds" of the image for the blur kernel to work on. Also, because one has kind of created and shifted these multiple planes of image data up/down/left/right, your image extents will now be larger. This is really easy to see with a simple box blur (each sample), but this is basically what you're seeing with the gaussian blur as well.
One easy route to a fix would be for you to just use the "blur" composition that is in your patch list (if you're using 10.6+). Also, If you have this, look for the source file at the location described in the patch list, and examine the process that is taken in that composition to properly setup the image to be cropped. If you follow that pattern for other core image kernels that have this kind of side effect, you will be pretty solid most of the time. Usually, one uses a crop patch referenced to the image dimensions of the source, and sometimes incorporates something like an affine clamp to extend the image if needed. On Thu, Sep 8, 2011 at 11:42 AM, Sebastian Mecklenburg < [email protected]> wrote: > Hello, > > OK, I'm getting closer. I have my shape generator up and running and it > looks fine. > However, I created my image just big enough to contain my shape, so when I > apply an image effect like "Blur" the image appears to be cropped. I think, > the image should be extended so it can fully contain the blur halo. Is this > something I have to take into account as a programmer or is this the > expected behavior and the user has to deal with it? I noticed that the > "Gaussian Blur" effect seems to shrink the image to be blurred beforehand to > make room for the blur halo, but I don't like that. > > Thanks and regards, > Sebastian Mecklenburg > > > _______________________________________________ > Do not post admin requests to the list. They will be ignored. > Quartzcomposer-dev mailing list ([email protected]) > Help/Unsubscribe/Update your Subscription: > > http://lists.apple.com/mailman/options/quartzcomposer-dev/gtoledo3%40gmail.com > > This email sent to [email protected] > -- George Toledo [email protected] www.georgetoledo.com
_______________________________________________ Do not post admin requests to the list. They will be ignored. Quartzcomposer-dev mailing list ([email protected]) Help/Unsubscribe/Update your Subscription: http://lists.apple.com/mailman/options/quartzcomposer-dev/archive%40mail-archive.com This email sent to [email protected]
