which space pixelDepth in?

[douysu]

hi, thanks for your code. is good.

First question

I was confused the pixelDepth space in file filter-narrow-range.fs.glsl

void main()
{
    float pixelDepth = texture(u_DepthTex, f_TexCoord).r;

    if(pixelDepth > 0.0 || pixelDepth < -1000.0f) {
        outDepth = pixelDepth;
    } else {
        outDepth = (u_DoFilter1D == 1) ? filter1D(pixelDepth) : filter2D(pixelDepth);
    }
}

Which space is it? I think is clip Space by look the file depth-pass.fs.glsl code

    vec4 clipSpacePos = projectionMatrix * vec4(fragPos, 1.0);
    outDepth = fragPos.z;

and you limit the pixelDepth range on [-1000, 0], NDC space depth is [-1, 1].

Second question

In file composition-pass.fs.glsl, zn initial value is clip space? The purpose of this formula is convert eyespace to NDC space?

vec3 uvToEye(vec2 texCoord, float eyeDepth)
{
    float x  = texCoord.x * 2.0 - 1.0;
    float y  = texCoord.y * 2.0 - 1.0;
    float zn = ((farZ + nearZ) / (farZ - nearZ) * eyeDepth + 2 * farZ * nearZ / (farZ - nearZ)) / eyeDepth;

    vec4 clipPos = vec4(x, y, zn, 1.0f);
    vec4 viewPos = invProjectionMatrix * clipPos;
    return viewPos.xyz / viewPos.w;
}

Best.

[ttnghia]

Hi douysu,

Thank you for your interest and I'm sorry for the confusion---I didn't have much time to document carefully.

1. The depth value is in camera space, thus it can have any value from [-inf, inf]. Here we clip to keep the value within the view frustum.
2. The purpose of the uvToEye function is to convert back the tuple {texCoord, eyeDepth} back to 3D space (in eye coordinates). For the z_n value, you can have a look at its formula here (near the end of that page): http://www.songho.ca/opengl/gl_projectionmatrix.html

Hope that I answered your questions. Just let me know if you still have anything unclear.

[douysu]

Thanks for your answer.
file depth-pass.fs.glsl will produce the depth map and the value is outDepth, because is in camera space. But why also need the line code gl_FragDepth = clipSpacePos.z / clipSpacePos.w?

    out float outDepth;
    vec4 clipSpacePos = projectionMatrix * vec4(fragPos, 1.0);
    outDepth = fragPos.z;

    gl_FragDepth = clipSpacePos.z / clipSpacePos.w;

[ttnghia]

We need to output the depth value in eye space (outDepth) for further computation (filtering), and depth value in NDC space (gl_FragDepth) for depth test.

[douysu]

ok. thanks

[douysu]

hi, ttnghia. In file depth-pass.fs.glsl, clipSpacePos.z / clipSpacePos.w is NDC space and it's range is [-1, 1]. But gl_FragDepth range is [0, 1]. Why is gl_FragDepth = clipSpacePos.z / clipSpacePos.w; rather than gl_FragDepth = （clipSpacePos.z / clipSpacePos.w) * 0.5f + 0.5f;

Best~

[ttnghia]

I can't remember why I used that. Maybe it is a bug---I took it from somewhere else, and maybe mistakenly used it without fixing it. You can try your new formula for computing gl_FragDepth to see if it works 😄 .