[source lang="cpp"]float3x3 ComputeTangentFrame(float3 N,float3 p,float2 uv)
{
float3 dp1 = ddx(p);
float3 dp2 = ddy(p);
float2 duv1 = ddx(uv);
float2 duv2 = ddy(uv);
float3x3 M = float3x3(dp1,dp2,cross(dp1,dp2));
float2x3 inversetransposeM =
float2x3(cross(M[1],M[2]),cross(M[2],M[0]));
float3 T = mul(float2(duv1.x,duv2.x),inversetransposeM);
float3 B = mul(float2(duv1.y,duv2.y),inversetransposeM);
return float3x3(normalize(T),normalize(B),N);
}[/source]
It seems good and simple,but soon I get punished at little seams with UV mirrors.The whole model looks right under light,but at those seams,normal gets wrong and light reflect in unproper way,so a sudden change or miss lighting on surface appears.I think this is because when current pixel is shared by different triangles with different directions of UV,the ddx and ddy instruction will get wrong answers.
So,I use
[source lang="cpp"]float r = (dot(cross(T,B),N)<0.0f) ? -1.0f : 1.0f;[/source]
to judge whether UV was mirrored.Then I realized I can't change tangent or binormal or normal,for most part of the mesh already seems right.On the other hand,I cannot know whether it's U mirror or V mirror.It's hard to tell the right direction of two axes(tangent and binormal) when you only know one axes(Normal).The author do not give futher information.
Btw,because I'm using dx9,so I can't split vertexes in shader.
So I want to know is there a possible way to solve this problem?Thanks for any reply
