Hey everybody,
I was searching for an answer to this question on Google and it seems like you guys know a lot about advanced OpenGL techniques, so maybe someone can help!
I'm working on drawing app that uses OpenGL ES 1.1, and I'd like to implement photoshop-style layers (where each layer in the image is drawn over the previous layer with varying opacity.)
I've gotten it to work using a texture for each layer and compositing with glBlend(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA), but I need to optimize the drawing code by taking all of the layers above the current one and writing them into a single texture that can then be drawn quickly. So, for example:
-- Texture --
-- Texture --
-- Texture --
-- Texture User is Drawing Into --
-- Texture --
-- Texture --
-- Texture --
becomes
-- "Flattened" Texture of 3 Textures --
-- Texture User is Drawing Into --
-- "Flattened" Texture of 3 Textures --
That way, only three textures need to be drawn to the screen with each refresh. The problem is, I'm having trouble preserving alpha values with the optimized approach. In both cases, I'm using glBlend(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA), but the result is not the same. In the optimized approach, drawing the second "3 layer texture" seems to draw 50% white in regions that should be transparent.
Do you know if this approach is feasible? Is there a way I can combine textures and maintain alpha values? glBlend seems to do things nicely, but the destination alpha doesn't seem to be source alpha + destination alpha, it seems to be 1.0 all the time...
Here's the source for the two approaches I have now:
Basic Approach (this works fine!)
glBlend(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
glBindTexture(GL_TEXTURE_2D, layer1Texture);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
glBindTexture(GL_TEXTURE_2D, layer2Texture);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
glBindTexture(GL_TEXTURE_2D, layer3Texture);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
glBindTexture(GL_TEXTURE_2D, layer4Texture);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
glBindTexture(GL_TEXTURE_2D, layer5Texture);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
Optimized Approach:
// this code is run when when the current "active" layer changes. Two textures
// "belowTexture" and "aboveTexture" are filled with the contents of the layers
// below and above the current one, respectively. Just uses
// glFramebufferTexture2DOES and then draws using the same approach as above..
// bind the "belowTexture"
glBindFramebufferOES(GL_FRAMEBUFFER_OES, compositingFramebuffer);
glFramebufferTexture2DOES(GL_FRAMEBUFFER_OES, GL_COLOR_ATTACHMENT0_OES,GL_TEXTURE_2D, belowTexture, 0);
// clear it
glClearColorx(0,0,0,0);
glClear(GL_COLOR_BUFFER_BIT);
// prepare for drawing textures...
glEnableClientState(GL_TEXTURE_COORD_ARRAY);
glEnable(GL_TEXTURE_2D);
glEnable(GL_BLEND);
glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
// draw textures (it would be a loop)
glBindTexture(GL_TEXTURE_2D, layer1Texture);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
glBindTexture(GL_TEXTURE_2D, layer2Texture);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
// Then when we go to actually draw:
if (currentLayer > 0){
glBlendFunc(GL_ONE, GL_ZERO);
glBindTexture(GL_TEXTURE_2D, belowTexture);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
}
glBlendFunc(GL_SRC_ALPHA, GL_ONE_MINUS_SRC_ALPHA);
glBindTexture(GL_TEXTURE_2D, currentLayerTexture);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
if (currentLayer < [layers count]){
glBindTexture(GL_TEXTURE_2D, aboveTexture);
glDrawArrays(GL_TRIANGLE_STRIP, 0, 4);
}
Any help would be really appreciated! Hopefully I will get good enough at this to contribute to these forums :-)