Gbuffer的法向量的压缩

Posted 2021-08-01 潘李亮

BTW： 自己看到了记录一下，大家随意。

原文链接：https://aras-p.info/texts/CompactNormalStorage.html

1.直接存在RGB通道上。

half4 encode(half3 n){return half4(n.xyz*0.5+0.5,0);}
half3 decode(half4 c){return c.rgb*2-1;}

2.存储法线的XY在颜色通道上，计算Z

half4 encode(half3 n){return half4(n.xy*0.5+0.5,0.0);}
half3 decode(half2 c)
{
    half3 n;
    n.xy = c*2-1;
    n.z = sqrt(1-dot(n.xy,n.xy));
    return n;
}

优点：容易实现

缺点：误差较大。

3.球面坐标（Spherical Coordinates）

#define kPI 3.1415926536f
half4 encode (half3 n)
{
    return half4((half2(atan2(n.y,n.x)/kPI, n.z)+1.0)*0.5,0,0);
}
half3 decode (half2 enc)
{
    half2 ang = enc*2-1;
    half2 scth;
    sincos(ang.x * kPI, scth.x, scth.y);
    half2 scphi = half2(sqrt(1.0 - ang.y*ang.y), ang.y);
    return half3(scth.y*scphi.x, scth.x*scphi.x, scphi.y);
}

优点：适合大部分法线，不一定需要view space

缺点：使用了三角函数，计算量比较大。

4.Spheremap Transform 

half2 encode (half3 n)
{
    half2 enc = normalize(n.xy) * (sqrt(-n.z*0.5+0.5));
    enc = enc*0.5+0.5;
    return enc;
}
half3 decode (half4 enc)
{
    half4 nn = enc*half4(2,2,0,0) + half4(-1,-1,1,-1);
    half l = dot(nn.xyz,-nn.xyw);
    nn.z = l;
    nn.xy *= sqrt(l);
    return nn.xyz * 2 + half3(0,0,-1);
}

优点：效果还不错，计算量小。CryEngine 3使用的方案。 presented by Martin Mittring in "A bit more Deferred" presentation

5.Lambert Azimuthal Equal-Area projection

half2 encode (half3 n)
{
    half f = sqrt(8*n.z+8);
    return n.xy / f + 0.5;
}
half3 decode (half4 enc)
{
    half2 fenc = enc*4-2;
    half f = dot(fenc,fenc);
    half g = sqrt(1-f/4);
    half3 n;
    n.xy = fenc*g;
    n.z = 1-f/2;
    return n;
}

优点：效果还不错，计算量小。 by Sean Barrett

6.Stereographic Projection 

half4 encode (half3 n)
{
    half scale = 1.7777;
    half2 enc = n.xy / (n.z+1);
    enc /= scale;
    enc = enc*0.5+0.5;
    return half4(enc,0,0);
}

half3 decode (half4 enc)
{
    half scale = 1.7777;
    half3 nn =
        enc.xyz*half3(2*scale,2*scale,0) +
        half3(-scale,-scale,1);
    half g = 2.0 / dot(nn.xyz,nn.xyz);
    half3 n;
    n.xy = g*nn.xy;
    n.z = g-1;
    return n;
}

优点：效果还不错，计算量小。

8. Per-pixel View Space

float3x3 make_view_mat (float3 view)
{
    view = normalize(view);
    float3 x,y,z;
    z = -view;
    x = normalize (float3(z.z, 0, -z.x));
    y = cross (z,x);
    return float3x3 (x,y,z);
}

half4 encode (half3 n, float3 view)
{
    return half4(mul (make_view_mat(view), n).xy*0.5+0.5,0,0);
}

half3 decode (half4 enc, float3 view)
{
    half3 n;
    n.xy = enc*2-1;
    n.z = sqrt(1+dot(n.xy,-n.xy));
    n = mul(n, make_view_mat(view));
    return n;
}

优点：效果还不错，计算量小。

性能对比：

GPU performance comparison in a single table:

	#1: X & Y	#3: Spherical	#4: Spheremap	#7: Stereo	#8: PPView
Encoding, GPU cycles
Radeon HD2400	1.00	17.00	3.00	4.00	11.00
Radeon HD5870	0.50	0.95	0.50	0.50	0.80
GeForce 6200	1.00	12.00	4.00	2.00	12.00
GeForce 8800	7.00	43.00	12.00	12.00	24.00
Decoding, GPU cycles
Radeon HD2400	1.00	17.00	3.00	4.00	11.00
Radeon HD5870	0.50	0.95	0.50	1.00	0.80
GeForce 6200	4.00	7.00	6.00	4.00	12.00
GeForce 8800	15.00	23.00	15.00	12.00	29.00
Encoding, D3D ALU+TEX instruction slots
SM3.0	1	26	4	5	17
Decoding, D3D ALU+TEX instruction slots
SM3.0	8	18	9	8	22

质量对比

Quality comparison in a single table. PSNR based, higher numbers are better.

Method	PSNR, dB
#1: X & Y	18.629
#3: Spherical	42.042
#4: Spheremap	48.071
#7: Stereographic	44.147
#8: Per pixel view	38.730