How can I get depth value from depth texture?

Question

I have complete-filled depth texture, but it seems not to get a proper value from it.

GLSL code, fragment shader:

uniform sampler2D depth_tex;
uniform float viewport_w;
uniform float viewport_h;

void main()
{
    vec2 coord = vec2(gl_FragCoord.xy);
    float depth = texture2D(depth_tex, vec2(coord.x / viewport_w, coord.y / viewport.h)).r;
    float c = material_amb * depth; // material_amb is preset by host
    out_color = vec4(c, c, c, 1.0f);
}

I expect model's color will be changed by vertices' depth, but I can't see everything. Of course, I can see model without multiplying depth. depth texture's type is GL_DEPTH_COMPONENT, GL_FLOAT. What's the problem? I think a solution to get a depth value from a texture is wrong, but I don't know the reason. Where is a depth value stored in? And How can I get a depth value?

A useful tip to debug this kind of stuff: rather than multiplying by depth directly, multiply by fract(depth), that way you can figure out if the values are outside the [0..1] range to begin with. — UncleZeiv
By the way, there's also a typo in your code, viewport.h should be viewport_h... maybe the shader is not compiling when you are using depth? — UncleZeiv

Nicol Bolas Nicol Bolas · Accepted Answer · 2011-08-30T10:41:07

depth texture's type is GL_DEPTH_COMPONENT, GL_FLOAT

This makes no sense. GL_FLOAT is not an OpenGL image format. GL_DEPTH_COMPONENT is an image format, but it is an unsized one. It specifies an unsigned, normalized integer depth format of an unspecified size.

GL_FLOAT is probably what you put in the parameter second from the end in your glTexImage call, right? That is the pixel transfer parameter; it describes the pixel data you're copying to the texture, not the way the pixel data is stored within the texture. OpenGL can convert between many different kinds of data formats.

If your depth data comes from rendering to the depth buffer, then you must understand how to convert from the non-linear post-projection space to a linear pre-projection space before you can do something meaningful with that data. This answer shows how to do the conversion.

How can I get depth value from depth texture?

1 Answers