germaine

Andrew
interesting to see what you do here

"andrew" <matseevsky@rambler.ru> schreef in bericht
newsp.t79i4mrkb7jv7v@macek...
>
> Usual methods for gradient fills are based on meshes- triangular,
> rectangular, distorted rectangular. People, the main purpose of these
> meshes is to split region to be painted. After it each cell will be
> painted, using three or four corner points. That's all. Sorrowful
> situation, when programmers reload they own task to poor users. Users have
> to make some absolutely odd ops (create meshes). Moreover, try to add one
> additional control point to existing mesh, and this mesh forces you to add
> a whole row an column, even if one additional control point would be
> enough. Method, what I name «advanced gradient fills», require only set of
> vertices and bounding curve. That's all. Nice and very effective thing,
> but it could be done better too. When I'm converting a photorealistic
> raster to vector, I have at least one problem. Not all contours, what
> present in a source, enclose something. Many of them are open arcs, and I
> have to create additional lines, what are absent in a source raster.
> Course, meshes requires much more additional contours, but perfect method
> shouldn't use such kind of objects at all. Well, there is a way to do
> without them. This method, in fact, isn't finished now. But, some
> intermediate results looks rather promising. Two images are attached here.
> Two curves at each image, none of them splits bounding rectangular. Hope,
> later I'll be able to show, what can be done with real raster, using this
> technology. (Intel will kiss me, because this method require lots of
> computations).
> Andrew.

andrew

22.03.08 в 06:01 germaine в �воём пи�ьме пи�ал(а):

> Andrew
> interesting to see what you do here
>
> "andrew" <matseevsky@rambler.ru> schreef in bericht
> newsp.t79i4mrkb7jv7v@macek...
>>

Look at my next topic, what I've placed yesterday...