I have a 24-bit bitmaps with R, G, B color channels and a 24-bit bitmap with R, G, B alpha channels. I want to alpha blend the first bitmap to a HDC in GDI or RenderTarget in Direct2D with the alpha channels respectively. For example, suppose for one pixel, the bitmap color is (192, 192, 192), the HDC color is (0, 255, 255) and the alpha channels are (30, 40, 50). The final HDC color should be (22, 245, 242). I know I can BitBlt the HDC to a memory HDC first, do alpha blending by manually calculating the color of each pixel and finally BitBlt back. I just want to avoid the additional blitting and leave APIs do their job (faster since they are in kernel space).

The first idea comes to my mind is to split the source bitmap into 3 red-only, green-only and blue-only 8-bit bitmaps, do normal alpha blending, then composite the 3 output bitmaps into the HDC. But I don't find a way to do the splitting and composition natively in Windows (would Direct2D layer help?). Also, the splitting and compositing may require many additional copying. The performance overhead would be too high.

Or maybe do the alpha blending in 3 passes. Each pass apply the blending for one channel, while maintaining the other 2 unchanged.

Thanks for any comment.

EDIT: I found this question, and the answer should be good reference to this problem. However, besides AC_SRC_OVER, there is no other blending operation supported. Why don't Microsoft improve their API?