在UnrealEngine中用Custom节点实现高斯模糊

3x3高斯模糊
//input sW 分辨率宽
//input sH 分辨率高
//input NotUse 为了开启SceneTextureLookup函数而连接的节点，但是不参与逻辑
//input UV 屏幕缓存的坐标
//14是原始颜色
int tIndex=14;
float4 x1y1=0.0625*SceneTextureLookup(UV+float2(-1.0f/sW,-1.0/sH),tIndex,false);
float4 x2y1=0.125*SceneTextureLookup(UV+float2(0.0f,-1.0/sH),tIndex,false);
float4 x3y1=0.0625*SceneTextureLookup(UV+float2(1.0f/sW,-1.0/sH),tIndex,false);
float4 x1y2=0.125*SceneTextureLookup(UV+float2(-1.0f/sW,0.0f),tIndex,false);
float4 x2y2=0.25*SceneTextureLookup(UV+float2(0.0f,0.0f),tIndex,false);
float4 x3y2=0.125*SceneTextureLookup(UV+float2(1.0f/sW,0.0f),tIndex,false);
float4 x1y3=0.0625*SceneTextureLookup(UV+float2(-1.0f/sW,1.0/sH),tIndex,false);
float4 x2y3=0.125*SceneTextureLookup(UV+float2(0.0f,1.0/sH),tIndex,false);
float4 x3y3=0.0625*SceneTextureLookup(UV+float2(1.0f/sW,1.0/sH),tIndex,false);
return x1y1+x2y1+x3y1+x1y2+x2y2+x3y2+x1y3+x2y3+x3y3;

　　

3X3均值模糊
//input sW 分辨率宽
//input sH 分辨率高
//input NotUse 为了开启SceneTextureLookup函数而连接的节点，但是不参与逻辑
//input UV 屏幕缓存的坐标
//14是原始颜色
int tIndex=14;
float4 x1y1=SceneTextureLookup(UV+float2(-1.0f/sW,-1.0/sH),tIndex,false);
float4 x2y1=SceneTextureLookup(UV+float2(0.0f,-1.0/sH),tIndex,false);
float4 x3y1=SceneTextureLookup(UV+float2(1.0f/sW,-1.0/sH),tIndex,false);
float4 x1y2=SceneTextureLookup(UV+float2(-1.0f/sW,0.0f),tIndex,false);
float4 x2y2=SceneTextureLookup(UV+float2(0.0f,0.0f),tIndex,false);
float4 x3y2=SceneTextureLookup(UV+float2(1.0f/sW,0.0f),tIndex,false);
float4 x1y3=SceneTextureLookup(UV+float2(-1.0f/sW,1.0/sH),tIndex,false);
float4 x2y3=SceneTextureLookup(UV+float2(0.0f,1.0/sH),tIndex,false);
float4 x3y3=SceneTextureLookup(UV+float2(1.0f/sW,1.0/sH),tIndex,false);
return (x1y1+x2y1+x3y1+x1y2+x2y2+x3y2+x1y3+x2y3+x3y3)/9;

　　看了官方论坛的帖子，法线Custom里面可以写for循环，于是……

//5x5高斯模糊
//input sW 分辨率宽
//input sH 分辨率高
//input NotUse 为了开启SceneTextureLookup函数而连接的节点，但是不参与逻辑
//input UV 屏幕缓存的坐标
//14是原始颜色
int tIndex=14;
int UVOfferset[]={-2,-1,0,1,2};
float Weights[]=
{
	0.0036,0.0146,0.0256,0.0146,0.0036,
	0.0146,0.0586,0.0952,0.0586,0.0146,
	0.0256,0.0952,0.1501,0.0952,0.0256,
	0.0146,0.0586,0.0952,0.0586,0.0146,
	0.0036,0.0146,0.0256,0.0146,0.0036
};
float3 OutColor={0.0,0.0,0.0};
for(int i=0;i<=4;i++)
{
		for(int j=0;j<=4;j++)
		{
			OutColor+=Weights[i*5+j]*SceneTextureLookup(UV+float2(UVOfferset[j]/sW,UVOfferset[i]/sH),tIndex,false).xyz;
		}
}
return float4(OutColor,1.0f);

　　

参考了一下网上的文章写了个可以动态调整的高斯模糊

//动态高斯模糊
//input sW 分辨率宽
//input sH 分辨率高
//input NotUse 为了开启SceneTextureLookup函数而连接的节点，但是不参与逻辑
//input Radius 模糊半径
//input UV 屏幕缓存的坐标
//14是原始颜色
int tIndex=14;
if(Radius<1.0)
	return SceneTextureLookup(UV,tIndex,false);
float3 OutColor={0.0,0.0,0.0};
Radius=floor(Radius);
float Sigma=Radius/3;
float Sigma2=2*Sigma*Sigma;
int Number=Radius;
float WeightsSum;
for(int i=-Number;i<=Number;i++)
{
	for(int j=-Number;j<=Number;j++)
	{
		float Weights=(1/(Sigma2*3.1415))*exp(-(j*j+i*i)/Sigma2);
		WeightsSum+=Weights;
		OutColor+=Weights*SceneTextureLookup(UV+float2(j/sW,i/sH),tIndex,false).xyz;
	}
}
OutColor/=WeightsSum;
return float4(OutColor,1.0f);

　　

群里朋友需要一个模糊贴图的于是又写了个

//5x5高斯模糊
//input UV 贴图UV
//input Texture 传入TextureObject
//input TexSize 图片大小，比如float2（512,512）
int UVOfferset[]={-2,-1,0,1,2};
float Weights[]=
{
    0.0036,0.0146,0.0256,0.0146,0.0036,
    0.0146,0.0586,0.0952,0.0586,0.0146,
    0.0256,0.0952,0.1501,0.0952,0.0256,
    0.0146,0.0586,0.0952,0.0586,0.0146,
    0.0036,0.0146,0.0256,0.0146,0.0036
};
float3 OutColor={0.0,0.0,0.0};
for(int i=0;i<=4;i++)
{
        for(int j=0;j<=4;j++)
        {
			OutColor+=Weights[i*5+j]*Texture2DSample(Texture, TextureSampler, UV+float2(UVOfferset[j]/TexSize.x,UVOfferset[i]/TexSize.y));
        }
}
return float4(OutColor,1.0f);

　　

//7x7均值模糊
//input UV 贴图UV
//input Texture 传入TextureObject
//input TexSize 图片大小，比如float2（512,512）
int UVOfferset[]={-3,-2,-1,0,1,2,3};
float Weights[]=
{
	1,1,1,1,1,1,1,
	1,1,1,1,1,1,1,
	1,1,1,1,1,1,1,
	1,1,1,1,1,1,1,
	1,1,1,1,1,1,1,
	1,1,1,1,1,1,1,
	1,1,1,1,1,1,1
};
float3 OutColor={0.0,0.0,0.0};
for(int i=0;i<=6;i++)
{
        for(int j=0;j<=6;j++)
        {
			OutColor+=Weights[i*7+j]*Texture2DSample(Texture, TextureSampler, UV+float2(UVOfferset[j]/TexSize.x,UVOfferset[i]/TexSize.y));
        }
}
return float4(OutColor,1.0f)/49;