Question

I’m drawing 2D, concave, sometimes multicontoured, sometimes self intersecting polygons with OpenGL.
Here is a sample:

Right now, I take the points which if connected would result in the polygon’s outline. Then I put these into the GLUTesselator where triangles come out. I then make texture coordinates and texture the polygon.

The absolute slowest component is the tessellation / triangulation. Given that I just need to draw these as you see, what are faster alternatives to triangulating? Could I possibly find a painting algorithm that would set the pixel appropriately?

Thanks

Answer 1

First a simpler problem: suppose you just wanted to use one color. You could start with the list of outlines, then scan the whole window by row and column: whenever you crossed a boundary, you would increment or decrement (depending on the sense of the crossing) a counter how_many_outlines_am_I_inside. When it’s zero, paint the pixel white, otherwise paint it green (and I guess black for the boundary itself). This will handle self-intersecting curves correctly.

Now for the shades. Judging by your example, shading is uniform vertically, but changes horizontally, scaled to the full width of the outline. So instead of a simple counter, you’ll need a stack (I recommend stl::list) of outlines, so that you can keep track of which outline (of the ones you’re inside) is uppermost, so that you can calculate what fraction of the distance you’ve covered from x_min to x_max. Finally for the translucency (as with the star partly visible through the rectangle): you’ll have to decide on the rules yourself, I can’t infer them by eyeball, but the stack should make it easy to implement.

Want some help with the C++ code?