[whatwg] Adding SVG Filter-like functionality to Canvas 2D Context

[hansschmucker at gmail.com]

Hi everybody.

I've recently done some experiments using SVG filters (see  
http://www.tapper-ware.net/stable/web.filter.voxels/index.xhtml ). SVG  
Filters basically offer greater speed for users and easier optimization for  
implementing parties than trying to implement standard image manipulation  
through interpreted or compiled Javascript code. Filter operations can also  
easily be implemented via graphics accelerators, which results in far  
greater speed, which is especially important for mobile devices.

A large share of the typical operations performed by Canvas developers can  
be expressed easier via an SVG Filter like interface. Like contrast,  
brightness, blur and so on. Right now, such operations can sometimes be  
done in a hackish way, by combining the target image with a solid  
black/white picture and then masking it, or using getImageData, which is  
slow.

Mozilla also includes drawWindow, which allows a filtered element to be  
imported back into the Canvas, but there are dozens of security issues if  
you try to do that on the web (which is why it's only allowed for chrome).  
Aside from drawWindow, we are currently unable to build a processing chain  
that does include filters, other than applying a filter to the result  
rendered by the canvas via foreignObject.

SVG Filters are a relatively easy spec, where the most important parts can  
be implemented in a matter of hours. Also, since all browsers that  
currently support Canvas also support SVG, the actual functionality is  
already included in the source, so very little new code would have to be  
written.

I'm willing to write a spec for it if there's any interest, but for now,  
I'll just give you an example of how this could work.

The 2D Context gets a new method "createFilterChain()", that if invoked  
returns a Canvas2DContextFilterChain object.
A Canvas2DContextFilterChain is applied to any drawing much like  
globalAlpha via a globalFilterChain property.
A Canvas2DContextFilterChain is bound to a specific Canvas and can not be  
applied to anything else.
The affected region is by default the rectangle of the operation.
Canvas2DContextFilterChain instances have methods to add  
Canvas2DContextFilterNode elements, one for each type.
Note that there is no feImage equivalent to keep security complexities to a  
minimum.
Data is passed into a filter chain as BackgroundImage/Alpha (containing the  
image data in the target rectangle) and SourceImage/SourceAlpha for the  
data that is supposed to be drawn.
Each Canvas2DContextFilterNode operates in the RGBA32 space, no other color  
space is supported.
By default, each Canvas2DContextFilterNode is applied to a 0%,0% 100%,100%  
rectangle (the -10%,-10% rule of SVG is not applied)
For each Node, the coordinate system can be set to be Canvas relative or  
affected region relative.
Units are either numbers for absolute offsets from the chosen rectangle or  
percentages of the same.
Instead of the string based buffer system of SVG, each  
Canvas2DContextFilterNode has connectOutputImage/Alpha methods.


Am I the only one seeing any benefit for this or does anybody else think  
that would be hope for such a proposal?

--hans
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.whatwg.org/pipermail/whatwg-whatwg.org/attachments/20090706/c11d9b69/attachment.htm>