Logarithmic depth buffer

Question

I use a logarithmic depth buffer: in my shader I wrote the code describes here.

In the vertex shader:

void main()
{
  vec4 vertexPosClip = gl_ModelViewProjectionMatrix * gl_Vertex;
  gl_Position = vertexPosClip;
  gl_TexCoord[6] = vertexPosClip;
}

In the fragment shader:

void main()
{
  gl_FragColor = ...
  const float C = 1.0;
  const float far = 1000000000.0;
  const float offset = 1.0;
  gl_FragDepth = (log(C * gl_TexCoord[6].z + offset) / log(C * far + offset));
}

Now in one of my algorithms (SSAO which works well with a classic depth buffer), I want to recover the depth of my fragment. Here is the formula I use:

    const float C = 1.0;
    const float far = 1000000000.0;
    const float offset = 1.0;
//depth
    z=(pow(C*far+offset,depth_in_depth_buffer)-offset)/C;

However it doesn't work, the algorithm doesn't work anymore. Do you know if the z I have computed is the correct depth?

Answer 1

How did you prepare the gl_ModelViewProjectionMatrix? If you used the standard one, vertexPosClip.z will contain depth that is already projected, and thus wrong.

Without modifying the matrix, you can use the w component that contains positive depth into the scene:

gl_FragDepth = (log(C * gl_TexCoord[6].w + offset) / log(C * far + offset));

The computation of z from the depth buffer value seems to be correct, just don't forget you are getting a positive depth into the scene, not a negative one as is standard in OpenGL.