- From: Jim Mork <jmork@jasc.com>
- Date: Thu, 13 Jul 2000 15:35:01 -0500
- To: "'www-svg@w3.org'" <www-svg@w3.org>
Jon: >> If provided, filterRes indicates exactly how many pixels in X and Y are used for the intermediate offscreen bitmaps. It thus defines a particular resolution at which the filters are performed. This ability was deemed necessary to give authors the ability to ensure consistent results across different media for the filter effects, which might have different visual results depending on resolution. If not provided, then the user agent picks a suitable resolution appropriate to the device. Thanks, Jon. Maybe I have missed something obvious, but one of the general problems I'm not clear on is how to efficiently handle issues arising due to the non-local nature of some filter primitives--specifically, feConvolveMatrix, feDisplacementMap, feGaussianBlur, feMorphology and feOffset. When our viewer/editor zooms in on a graphic with a filter effect applied to it, what are we supposed to do? To illustrate the problem imagine the following scenario. Let's say we have a 100 x 100 pixel graphic with a Gaussian Blur applied to it. We want to zoom in a factor of 10. To display this properly, we will have to expand the blur radius by a factor of ten as well. With high zoom levels and large graphics, it won't take long for this to become extremely computationally expensive, to say the least. Even if the zoomed graphic is so large little of it fits in the view, that is little consolation because of the non-local nature of the said primitives. In the case of zooming, an alternative is to render the graphic at 1:1 and resample it up to the zoomed level--but this technique seems to subvert the idea of "scalable" graphics. Perhaps I have missed something obvious, but the answer is not clear to me. Any advice will be greatly appreciated. Best regards, James Mork Senior Software Engineer Jasc Software, Inc.
Received on Thursday, 13 July 2000 16:38:18 UTC