Using Depth Textures

It is possible to create Render Textures where each pixel contains a high-precision Depth value. This is mostly used when some effects need the Scene’s Depth to be available (for example, soft particles, screen space ambient occlusion and translucency would all need the Scene’s Depth). Image Effects often use Depth Textures too. 它可以创建渲染纹理，其中每个像素包含一个高精度的深度值。这是主要用于当有些影响需要场景的深度可用（例如，软颗粒，屏幕空间环境光遮蔽和透明度都需要场景的深度）。图像效果经常使用深度纹理。

Pixel values in the Depth Texture range between 0 and 1, with a non-linear distribution. Precision is usually 32 or 16 bits, depending on configuration and platform used. When reading from the Depth Texture, a high precision value in a range between 0 and 1 is returned. If you need to get distance from the Camera, or an otherwise linear 0–1 value, compute that manually using helper macros (see below). 像素值在深度纹理范围在0和1之间，具有非线性分布。精度通常是32或16位，这取决于配置和使用的平台。当从深度纹理读取时，返回在0和1之间的范围内的高精度值。如果您需要从相机，或其他线性0 - 1值的距离，计算手动使用辅助宏（见下文）。

Depth Textures are supported on most modern hardware and graphics APIs. Special requirements are listed below: 深度纹理支持最现代的硬件和图形API。特殊要求如下所列：

Direct3D 11+ (Windows), OpenGL 3+ (Mac/Linux), OpenGL ES 3.0+ (Android/iOS), Metal (iOS) and consoles like PS4/Xbox One all support depth textures. Direct3D 9 (Windows) requires a graphics driver to support “INTZ” texture format to get depth textures. OpenGL ES 2.0 (iOS/Android) requires GL_OES_depth_texture extension to be present. WebGL requires WEBGL_depth_texture extension.

Depth Texture Shader helper macros

Most of the time, Depth Texture are used to render Depth from the Camera. The UnityCG.cginc include file contains some macros to deal with the above complexity in this case:

UNITY_TRANSFER_DEPTH(o): computes eye space depth of the vertex and outputs it in o (which must be a float2). Use it in a vertex program when rendering into a depth texture. On platforms with native depth textures this macro does nothing at all, because Z buffer value is rendered implicitly.

UNITY_OUTPUT_DEPTH(i): returns eye space depth from i (which must be a float2). Use it in a fragment program when rendering into a depth texture. On platforms with native depth textures this macro always returns zero, because Z buffer value is rendered implicitly.

COMPUTE_EYEDEPTH(i): computes eye space depth of the vertex and outputs it in o. Use it in a vertex program when not rendering into a depth texture.

DECODE_EYEDEPTH(i)/LinearEyeDepth(i): given high precision value from depth texture i, returns corresponding eye space depth.

Linear01Depth(i): given high precision value from depth texture i, returns corresponding linear depth in range between 0 and 1.

Note: On DX11/12, PS4, XboxOne and Metal, the Z buffer range is 1–0 and UNITY_REVERSED_Z is defined. On other platforms, the range is 0–1.

For example, this shader would render depth of its GameObjects:

Shader "Render Depth" {
    SubShader {
        Tags { "RenderType"="Opaque" }
        Pass {
            CGPROGRAM

            #pragma vertex vert
            #pragma fragment frag
            #include "UnityCG.cginc"

            struct v2f {
                float4 pos : SV_POSITION;
                float2 depth : TEXCOORD0;
            };

            v2f vert (appdata_base v) {
                v2f o;
                o.pos = UnityObjectToClipPos(v.vertex);
                UNITY_TRANSFER_DEPTH(o.depth);
                return o;
            }

            half4 frag(v2f i) : SV_Target {
                UNITY_OUTPUT_DEPTH(i.depth);
            }
            ENDCG
        }
    }
}

Camera’s Depth Texture

A Camera can generate a depth, depth+normals, or motion vector Texture. This is a minimalistic G-buffer Texture that can be used for post-processing effects or to implement custom lighting models (e.g. light pre-pass). It is also possible to build similar textures yourself, using Shader Replacement feature.

The Camera’s depth Texture mode can be enabled using Camera.depthTextureMode variable from script.

There are three possible depth texture modes:

DepthTextureMode.Depth: a depth texture.

DepthTextureMode.DepthNormals: depth and view space normals packed into one texture.*

DepthTextureMode.MotionVectors: per-pixel screen space motion of each screen texel for the current frame. Packed into a RG16 texture. These are flags, so it is possible to specify any combination of the above textures.

DepthTextureMode.Depth texture

This builds a screen-sized depth texture.

Depth texture is rendered using the same shader passes as used for shadow caster rendering (ShadowCaster pass type). So by extension, if a shader does not support shadow casting (i.e. there’s no shadow caster pass in the shader or any of the fallbacks), then objects using that shader will not show up in the depth texture.

Make your shader fallback to some other shader that has a shadow casting pass, or

If you’re using surface shaders, adding an addshadow directive will make them generate a shadow pass too. Note that only “opaque” objects (that which have their materials and shaders setup to use render queue <= 2500) are rendered into the depth texture.

DepthTextureMode.DepthNormals texture

This builds a screen-sized 32 bit (8 bit/channel) texture, where view space normals are encoded into R&G channels, and depth is encoded in B&A channels. Normals are encoded using Stereographic projection, and depth is 16 bit value packed into two 8 bit channels.

UnityCG.cginc include file has a helper function DecodeDepthNormal to decode depth and normal from the encoded pixel value. Returned depth is in 0..1 range.

For examples on how to use the depth and normals texture, please refer to the EdgeDetection image effect in the Shader Replacement example project or Screen Space Ambient Occlusion Image Effect.

DepthTextureMode.MotionVectors texture

This builds a screen-sized RG16 (16-bit float/channel) texture, where screen space pixel motion is encoded into the R&G channels. The pixel motion is encoded in screen UV space.

When sampling from this texture motion from the encoded pixel is returned in a rance of –1..1. This will be the UV offset from the last frame to the current frame.

Tips & Tricks

Camera inspector indicates when a camera is rendering a depth or a depth+normals texture.

The way that depth textures are requested from the Camera (Camera.depthTextureMode) might mean that after you disable an effect that needed them, the Camera might still continue rendering them. If there are multiple effects present on a Camera, where each of them needs the depth texture, there’s no good way to automatically disable depth texture rendering if you disable the individual effects.

When implementing complex Shaders or Image Effects, keep Rendering Differences Between Platforms in mind. In particular, using depth texture in an Image Effect often needs special handling on Direct3D + Anti-Aliasing.

In some cases, the depth texture might come directly from the native Z buffer. If you see artifacts in your depth texture, make sure that the shaders that use it do not write into the Z buffer (use ZWrite Off).

Shader variables

Depth textures are available for sampling in shaders as global shader properties. By declaring a sampler called _CameraDepthTexture you will be able to sample the main depth texture for the camera.

_CameraDepthTexture always refers to the camera’s primary depth texture. By contrast, you can use _LastCameraDepthTexture to refer to the last depth texture rendered by any camera. This could be useful for example if you render a half-resolution depth texture in script using a secondary camera and want to make it available to a post-process shader.

The motion vectors texture (when enabled) is avaialable in Shaders as a global Shader property. By declaring a sampler called ‘_CameraMotionVectorsTexture’ you can sample the Texture for the curently rendering Camera.

Under the hood

Depth textures can come directly from the actual depth buffer, or be rendered in a separate pass, depending on the rendering path used and the hardware. Typically when using Deferred Shading or Legacy Deferred Lighting rendering paths, the depth textures come “for free” since they are a product of the G-buffer rendering anyway.

When the DepthNormals texture is rendered in a separate pass, this is done through Shader Replacement. Hence it is important to have correct “RenderType” tag in your shaders.

When enabled, the MotionVectors texture always comes from a extra render pass. Unity will render moving GameObjects into this buffer, and construct their motion from the last frame to the current frame.