GPU Texture Splatting

Question

Just as a quick example, I'm trying to do the following:

---

+ + =

---

With the third image as an alpha map, how could this be implemented in a DX9-compatible pixel shader to "blend" between the first two images, creating an effect similar to the fourth image?
Furthermore, how could this newly created texture be given back to the CPU, where it could be placed back inside the original array of textures?

Answer 1

The rough way is to blend the colors of the textures with the alphamap and return it from the pixelshader:

float alpha = tex2D(AlphaSampler,TexCoord).r;
float3 texture1 = tex2D(Texture1Sampler,TexCoord).rgb;
float3 texture2 = tex2D(Texture2Sampler,TexCoord).rgb;
float3 color = lerp(texture1,texture2,alpha);
return float4(color.rgb,1);

Therefore you need a texture as rendertarget (doc) with the size of the inputtextures and a fullscreen quad as geometry for rendering, a xyzrhw quad would be the easiest. This texture you can use further for rendering. If you want to read the texels or something else, where you must lock the result you could work with StretchRect (doc) or UpdateSurface (doc) to copy the data into a normal texture.

If the performance isn't important (e.g. you preprocess the textures), you could easier compute this on the cpu (but it's slower). Lock the 4 textures, iterate over the pixels and merge them directly.