Graphics Reference
In-Depth Information
Figure 2.1.
The render targets for deferred rendering: from left to right, final image, normals, encoded
depth, and albedo color.
normal) information to a texture and then doing a second pass of geometry with
the diffuse textures to produce the final image. As this technique processes the
geometry twice, if the application is limited by vertex processing then this tech-
nique may not be the best option. However, if the application is limited by
fragment operations, then this technique may be a good option because it uses
less memory and therefore less memory bandwidth than other techniques.
OpenGL ES 2.0 and 3.0.
Light pre-pass rendering can be performed in both
OpenGL ES 2.0 and 3.0.
Limitations of both deferred and light pre-pass rendering include the inability
to do hardware antialiasing. A technique that allows this possibility is called light
indexed rendering. When using a depth texture, the application can avoid using
multiple render targets. Figure 2.2 shows the intermediate render targets for this
technique.
2.3.3 Light Indexed Rendering
Light indexed rendering [Treblico 09] is so-called because instead of storing the
required lighting information such as normals and color per pixel, we store an
index to the light affecting each pixel. We then use this index to sample separate
one-dimensional textures, each with a different property, i.e., light position and
light color. This acts like forward rendering; however, there is much more control
Figure 2.2.
Render targets for light pre-pass rendering: lighting texture (left), encoded depth (center),
and normals (right).
