AGG第四十二课 Blitting an image over another with transparency

问题：

I've managed to blit a loaded p_w_picpath onto another through the method

"copy_from(...)" of renderer_base.

I'd like to know how can i blit the same p_w_picpath and also specifying a color

that will NOT overwrite the pixels of the destination p_w_picpath ( a color that

will then result transparent ).

I complete my question by specifying that the source p_w_picpath is a BMP and

therefore doesn't support transparency and alpha channel in the pixel

values.

Any suggestion is appreciated

回答：

Yes, your simple question isn't actually that simple. I could add this

functionality to copy_from, but I don't know what exactly to add. Do you

need to make transparent only one color? Or a number of colors? Or maybe it

should be some volume in RGB with transparency determined by the distance

from a certain point. OK, suppose you need only one color, bu someone else

may want to make a number of colors translucent. And so on. We need a kind

of a general solution.

For now you can try to use a method similar to the one in

examples/p_w_picpath_alpha.cpp

See class span_conv_brightness_alpha_rgb8 and modify it as you want.

But it's a transformer, so that, works slower than copy_from. Besides, with

any kind of a filter (bilinear and such) the colors are not exact, so, it's

probably makes sense to use span_p_w_picpath_filter_rgb_nn instead of

span_p_w_picpath_filter_rgb_bilinear.

Maybe it makes sense to add some functor to copy_from(), but I have no idea

how to do that in a general way and without losing performance. Perhaps

there should be another function more like blend_from(). Any ideas?

答复：

Actually I've gone for a "specialized" solution, that just solves my case.

In detail, I created a new renderer that's only suitable to copy bitmaps one

over another with 1 transparent color.

I only have ported the "copy_from" method from renderer_base along with all

the clipping helper methods and obejcts of your original rendere_base.

I added a method that sets a trasparency color:

setTransaprentColor(color_type color)

{

...

}

In the "copy_from" method, I get the color of the source bitmap with the

"pixel(x,y)" method, check it against "transparentColor" and, if equal, I

simply skip the point with a "continue" statement.

Else, I call "copy_pixel" on the underlying rendering_buffer with the exact

value of the source bitmap pixel.

It's evident that it can be easily extended to a generic "array" of

transparent colors, with the cost of cycling through all of them to see if

one of them matches the source color. Even better, one could even imagine to

"map" some source color with some other colors (along with its own alpha

channel) once the source RGB values are matched against the RGB values of

the "transformer" object. I thought about it and it is not difficult to code

such a transformer object and pass it to a renderer. But that would be too

much for my special case and also kind of "expensive" in terms of

performance, so I just left it "for future expansions" ;-)