【Unity】2019 HDRP 7.3 CustomPass 高斯模糊_unity hdrp 图片 高斯模糊

2021HDRP模糊效果Demo git参考：

GitHub - alelievr/HDRP-Custom-Passes: A bunch of custom passes made for HDRP

1、创建HDRP项目 

2、创建CustomPass物体

新增2个FullScreenCustomPass

创建一个OutlineH材质球和OutlineH着色器，将OutlineH着色器拖拽给OutlineH材质球，OutlineH着色器代码如下：

Shader "Hidden/OutlineH"
{
	Properties{
		_BlurScale("BlurScale", float) = 1
	}
	HLSLINCLUDE
 
		#pragma vertex Vert
 
		#pragma target 4.5
		#pragma only_renderers d3d11 playstation xboxone vulkan metal switch
 
		#include "Packages/com.unity.render-pipelines.high-definition/Runtime/RenderPipeline/RenderPass/CustomPass/CustomPassCommon.hlsl"
	
		float _BlurScale;
 
		float2 ClampUVs(float2 uv)
		{
			uv = clamp(uv, 0, _RTHandleScale.xy - _ScreenSize.zw * 2); // clamp UV to 1 pixel to avoid bleeding
			return uv;
		}
 
		float2 GetSampleUVs(Varyings varyings)
		{
			float depth = LoadCameraDepth(varyings.positionCS.xy);
			PositionInputs posInput = GetPositionInput(varyings.positionCS.xy, _ScreenSize.zw, depth, UNITY_MATRIX_I_VP, UNITY_MATRIX_V);
			return posInput.positionNDC.xy * _RTHandleScale.xy;
		}
 
		float4 FullScreenPass(Varyings varyings) : SV_Target
		{
			UNITY_SETUP_STEREO_EYE_INDEX_POST_VERTEX(varyings);
 
			float depth = LoadCameraDepth(varyings.positionCS.xy);
			PositionInputs posInput = GetPositionInput(varyings.positionCS.xy, _ScreenSize.zw, depth, UNITY_MATRIX_I_VP, UNITY_MATRIX_V);
			float4 color = float4(0.0, 0.0, 0.0, 0.0);			
 
			// Load the camera color buffer at the mip 0 if we're not at the before rendering injection point
			if (_CustomPassInjectionPoint != CUSTOMPASSINJECTIONPOINT_BEFORE_RENDERING)
				color = float4(CustomPassSampleCameraColor(posInput.positionNDC.xy, 0), 1);		
 
			//UV获取方式,varyings.positionCS.xy * _ScreenSize.zw为NDC坐标,_RTHandleScale.xy是对NDC坐标进行一个缩放获取uv
			//float2 uv = ClampUVs(varyings.positionCS.xy * _ScreenSize.zw * _RTHandleScale.xy);
 
			//横向模糊（需要起另一个CustomPass处理纵向模糊）
			float3 sumColor = 0;
			float weight[3] = { 0.4026, 0.2442, 0.0545 };		
 
			sumColor = color.rgb * weight[0];
			for (int i = 1; i < 3; i++)
			{			
				//注意因为采样坐标是用NDC坐标,故float2(_BlurScale * i, 0) * _ScreenSize.zw是获取一个NDC坐标系上的增量(偏移值) 进行对周围横向或纵向像素采样!
				sumColor += CustomPassSampleCameraColor(posInput.positionNDC.xy + float2(_BlurScale * i, 0) * _ScreenSize.zw, 0).rgb * weight[i];
				sumColor += CustomPassSampleCameraColor(posInput.positionNDC.xy + float2(_BlurScale * -i, 0) * _ScreenSize.zw, 0).rgb * weight[i];
			}
		
			return float4(sumColor, 1);
		}
 
	ENDHLSL
 
	SubShader
	{
		Pass
		{
			Name "Custom Pass 0"
 
			ZWrite On
			ZTest Always
			Blend SrcAlpha OneMinusSrcAlpha 			
			Cull Off
 
			HLSLPROGRAM
				#pragma fragment FullScreenPass
			ENDHLSL
		}
	}
	Fallback Off
}

这一个材质效果是进行一次横向模糊，将结果缓存在CustomPass的缓冲区，默认是Camera（即直接输出到屏幕 但它还是会做一次混合的 仔细看Shader就能发现 Blend SrcAlpha OneMinusSrcAlpha）

新建一个OutlineV材质球和着色器，着色器代码：

Shader "Hidden/OutlineV"
{
	Properties{
		_BlurScale("BlurScale", float) = 1
	}
	HLSLINCLUDE
 
		#pragma vertex Vert
 
		#pragma target 4.5
		#pragma only_renderers d3d11 playstation xboxone vulkan metal switch
 
		#include "Packages/com.unity.render-pipelines.high-definition/Runtime/RenderPipeline/RenderPass/CustomPass/CustomPassCommon.hlsl"
	
		float _BlurScale;
 
		float2 ClampUVs(float2 uv)
		{
			uv = clamp(uv, 0, _RTHandleScale.xy - _ScreenSize.zw * 2); // clamp UV to 1 pixel to avoid bleeding
			return uv;
		}
 
		float2 GetSampleUVs(Varyings varyings)
		{
			float depth = LoadCameraDepth(varyings.positionCS.xy);
			PositionInputs posInput = GetPositionInput(varyings.positionCS.xy, _ScreenSize.zw, depth, UNITY_MATRIX_I_VP, UNITY_MATRIX_V);
			return posInput.positionNDC.xy * _RTHandleScale.xy;
		}
 
		float4 FullScreenPass(Varyings varyings) : SV_Target
		{
			UNITY_SETUP_STEREO_EYE_INDEX_POST_VERTEX(varyings);
 
			float depth = LoadCameraDepth(varyings.positionCS.xy);
			PositionInputs posInput = GetPositionInput(varyings.positionCS.xy, _ScreenSize.zw, depth, UNITY_MATRIX_I_VP, UNITY_MATRIX_V);
			float4 color = float4(0.0, 0.0, 0.0, 0.0);			
 
 
			//纵向从CustomPass(上一个横向模糊处理后的图片采样，即CustomColor缓冲区） 使用的是uv  
// float4 SampleCustomColor(float2 uv);
// float4 LoadCustomColor(uint2 pixelCoords);
 
			//UV获取方式,varyings.positionCS.xy * _ScreenSize.zw为NDC坐标,_RTHandleScale.xy是对NDC坐标进行一个缩放获取uv
			float2 uv = ClampUVs(varyings.positionCS.xy * _ScreenSize.zw * _RTHandleScale.xy);
 
			// Load the camera color buffer at the mip 0 if we're not at the before rendering injection point
			if (_CustomPassInjectionPoint != CUSTOMPASSINJECTIONPOINT_BEFORE_RENDERING)
				color = SampleCustomColor(uv);
						 
						
 
			//纵向向模糊
			float3 sumColor = 0;
			float weight[3] = { 0.4026, 0.2442, 0.0545 };		
 
			sumColor = color.rgb * weight[0];
			for (int i = 1; i < 3; i++)
			{			
				//注意采用使用的是UV坐标，因此float2(0, _BlurScale * i) * _ScreenSize.zw * _RTHandleScale.xy获取UV坐标增量（偏移值）
				sumColor += SampleCustomColor(uv + float2(0, _BlurScale * i) * _ScreenSize.zw * _RTHandleScale.xy).rgb * weight[i];				
				sumColor += SampleCustomColor(uv + float2(0, _BlurScale * -i) * _ScreenSize.zw * _RTHandleScale.xy).rgb * weight[i];
			}
		
			return float4(sumColor, 1);
		}
 
	ENDHLSL
 
	SubShader
	{
		Pass
		{
			Name "Custom Pass 0"
 
			ZWrite On
			ZTest Always
			Blend SrcAlpha OneMinusSrcAlpha 			
			Cull Off
 
			HLSLPROGRAM
				#pragma fragment FullScreenPass
			ENDHLSL
		}
	}
	Fallback Off
}

注意看选中的是Camera,第二次采样的是CustomPass的图片（第一次横向模糊输出的图片）对它进行了一个纵向模糊，然后输出到Camera。

代码很简单，比较难理解的是第一个横向模糊shader是使用

CustomPassSampleCameraColor(posInput.positionNDC.xy, 0) 获取了一个float3，其中采样坐标是NDC坐标。

第二个纵向模糊shader是使用

SampleCustomColor(uv) 获取float4 它是从CustomPass缓存的图片采样的，必须先进行横向模糊输出custom后 才能进行纵向模糊 ，当然你也可以反过来，使用先纵向后横向，但肯定是最后一个进行的模糊要是从Custom来获取上一次的处理结果，而不能再去使用CustomPassSampleCameraColor，因为这个看英文就知道它就是从摄像机缓存获取图片，这个就是当前渲染到的屏幕图而已。

重点内容是：

UV获取方式：varyings.positionCS.xy * _ScreenSize.zw为NDC坐标,_RTHandleScale.xy是对NDC坐标进行一个缩放获取uv

即 uv = varyings.positionCS.xy * _ScreenSize.zw * _RTHandleScale.xy);

uv = clamp(uv, 0, _RTHandleScale.xy - _ScreenSize.zw * 2); // clamp UV to 1 pixel to avoid bleeding

对uv进行clamp处理是将uv约束在1个像素范围以内，这个不太懂了。

其中NDC坐标系是用 varyings.positionCS.xy * _ScreenSize.zw 获取的，当你想做偏移float2(1,0)时，只要将这个float2(1,0)理解成positionCS.xy即可。

positonCS是Homogenous clip spaces（同质剪辑空间）裁剪空间？

_ScreenSize.zw是一个能让裁剪空间缩放回NDC坐标系的缩放系数？