openGLES 2 - How do I load a texture with 8-bit grayscale and 8-bit alpha, without expanding to 32-bit RGBA?

Question

We're trying to optimize memory usage in our program, and we've noticed that some of our textures could be stored as greyscale with an alpha value as when stored as RGBA the RGB values are all the same.

We'd like to save disk space as well as memory by only needing to store 2 channels: greyscale and alpha, both 8-bit. Thus providing a 50% saving on both disk and memory over 32-bit RGBA.

However the hurdle we've hit is that we can't work out how to give this info to openGL. We can store the image on disk either as one image with two 8-bit channels, or as two images with one 8-bit channel each. I can't find a way for openGL to see this as a greyscale image so that it only has these two channels in memory without looking at them as being specific colours like Red and Green - as I have found GL_RG8 exists which has the right number of channels at the right bitdepth, but wrong colours.

Is this even possible in openGLES 2? Can we give 2 channels to openGL and tell it that they're not Red and Green but Grey and Alpha?

I have seen this question: Can I use a grayscale image with the OpenGL glTexImage2D function? but that's rather old and GL_LUMINANCE is now depricated.

Answer 1

Note that GL_RG8 only exists on OpenGL ES 3.x upwards; it's not part of OpenGL ES 2.x. For ES 2.x you can use GL_LUMINANCE_ALPHA . It's deprecated, but not removed, so should still work just fine.

If you really want to use GL_RG8 then you can fix "wrong color" either by using swizzles in the shader code to reorder the channels when you use them, or you can swizzle in the sampler by setting TEXTURE_SWIZZLE_[R|G|B|A] using glTexParameteri() .

Answer 2

如果您不介意，我会注意到我遇到的每个ES2实现（iOS，Angle，Adreno，Mali和其他一些奇怪的Android驱动程序）都默默支持GL_RG8（用于格式）（和GL_RG用于internalFormat）。