opengl - depth buffer got by glReadPixels is always 1

Question

I'm using glReadPixels to get depth value of select pixel, but i always get 1, how can i solve it? here is the code:

    glEnable(GL_DEPTH_TEST);
    ..
    glReadPixels(x, viewport[3] - y, 1, 1, GL_DEPTH_COMPONENT, GL_FLOAT, z);

Do I miss anything? And my rendering part is shown below. I use different shaders to draw different part of scene, so how should i make it correct to read depth value from buffer?

void onDisplay(void)
{
// Clear the window and the depth buffer
glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);

// calculate the view matrix.
GLFrame eyeFrame;
eyeFrame.MoveUp(gb_eye_height);
eyeFrame.RotateWorld(gb_eye_theta * 3.1415926 / 180.0, 1.0, 0.0, 0.0);
eyeFrame.RotateWorld(gb_eye_phi * 3.1415926 / 180.0, 0.0, 1.0, 0.0);
eyeFrame.MoveForward(-gb_eye_radius);
eyeFrame.GetCameraMatrix(gb_hit_modelview);
gb_modelViewMatrix.PushMatrix(gb_hit_modelview);

// draw coordinate system
if(gb_bCoord)
{
    DrawCoordinateAxis();
}

if(gb_bTexture)
{

    GLfloat vEyeLight[] = { -100.0f, 100.0f, 150.0f };
    GLfloat vAmbientColor[] = { 0.2f, 0.2f, 0.2f, 1.0f };
    GLfloat vDiffuseColor[] = { 1.0f, 1.0f, 1.0f, 1.0f};

    glUseProgram(normalMapShader);
    glUniform4fv(locAmbient, 1, vAmbientColor);
    glUniform4fv(locDiffuse, 1, vDiffuseColor);
    glUniform3fv(locLight, 1, vEyeLight);
    glUniform1i(locColorMap, 0);
    glUniform1i(locNormalMap, 1);
    gb_treeskl.Display(SetGeneralColor, SetSelectedColor, 0);
}
else
{
    if(!gb_bOnlyVoxel)
    {
        if(gb_bPoints)
        {
            //GLfloat vPointColor[] = { 1.0, 1.0, 0.0, 0.6 };
            GLfloat vPointColor[] = { 0.2, 0.0, 0.0, 0.9 };
            gb_shaderManager.UseStockShader(GLT_SHADER_FLAT, gb_transformPipeline.GetModelViewProjectionMatrix(), vPointColor);
            gb_treeskl.Display(NULL, NULL, 1);
        }
        if(gb_bSkeleton)
        {
            GLfloat vEyeLight[] = { -100.0f, 100.0f, 150.0f };
            glUseProgram(adsPhongShader);
            glUniform3fv(locLight, 1, vEyeLight);
            gb_treeskl.Display(SetGeneralColor, SetSelectedColor, 0);
        }
    }
    if(gb_bVoxel)
    {
        GLfloat vEyeLight[] = { -100.0f, 100.0f, 150.0f };
        glUseProgram(adsPhongShader);
        glUniform3fv(locLight, 1, vEyeLight);
        SetVoxelColor();
        glPolygonMode(GL_FRONT, GL_LINE);
        glLineWidth(1.0f);
        gb_treeskl.DisplayVoxel();
        glPolygonMode(GL_FRONT, GL_FILL);
    }
}
//glUniformMatrix4fv(locMVP, 1, GL_FALSE, gb_transformPipeline.GetModelViewProjectionMatrix());
//glUniformMatrix4fv(locMV, 1, GL_FALSE, gb_transformPipeline.GetModelViewMatrix());
//glUniformMatrix3fv(locNM, 1, GL_FALSE, gb_transformPipeline.GetNormalMatrix());
//gb_sphereBatch.Draw();
gb_modelViewMatrix.PopMatrix();

glutSwapBuffers();

}

Answer 1

I think you are reading correctly the only problem is that you are not linearize the depth from buffer back to <znear...zfar> range hence the ~1 value for whole screen due to logarithmic dependence of depth (almost all the values are very close to 1).

I am doing this like this:

double glReadDepth(double x,double y,double *per=NULL)                  // x,y [pixels], per[16]
    {
    GLfloat _z=0.0; double m[16],z,zFar,zNear;
    if (per==NULL){ per=m; glGetDoublev(GL_PROJECTION_MATRIX,per); }    // use actual perspective matrix if not passed
    zFar =0.5*per[14]*(1.0-((per[10]-1.0)/(per[10]+1.0)));              // compute zFar from perspective matrix
    zNear=zFar*(per[10]+1.0)/(per[10]-1.0);                             // compute zNear from perspective matrix
    glReadPixels(x,y,1,1,GL_DEPTH_COMPONENT,GL_FLOAT,&_z);              // read depth value
    z=_z;                                                               // logarithmic
    z=(2.0*z)-1.0;                                                      // logarithmic NDC
    z=(2.0*zNear*zFar)/(zFar+zNear-(z*(zFar-zNear)));                   // linear <zNear,zFar>
    return -z;
    }

Do not forget that x,y is in pixels and (0,0) is bottom left corner !!! The returned depth is in range <zNear,zFar>. The function is assuming you are using perspective transform like this:

void glPerspective(double fovy,double aspect,double zNear,double zFar)
    {
    double per[16],f;
    for (int i=0;i<16;i++) per[i]=0.0;
    // original gluProjection
//  f=divide(1.0,tan(0.5*fovy*deg))
//  per[ 0]=f/aspect;
//  per[ 5]=f;
    // corrected gluProjection
    f=divide(1.0,tan(0.5*fovy*deg*aspect));
    per[ 0]=f;
    per[ 5]=f*aspect;
    // z range
    per[10]=divide(zFar+zNear,zNear-zFar);
    per[11]=-1.0;
    per[14]=divide(2.0*zFar*zNear,zNear-zFar);
    glLoadMatrixd(per);
    }

Beware the depth accuracy will be good only for close to camera object without linear depth buffer. For more info see:

• How to correctly linearize depth in OpenGL ES in iOS?

If the problem persist there might be also another reason for this. Do you have Depth buffer in your pixel format? In windows You can check like this:

• Getting a window's pixel format

Missing depth buffer could explain why the value is always 1 (not like ~0.997). In such case you need to change the init of your window enabling some bits for depth buffer (16/24/32). See:

• What is the proper OpenGL initialisation on Intel HD 3000?

For more detailed info about using this technique (with C++ example) see:

• OpenGL 3D-raypicking with high poly meshes