Hi!
I am in the early stages of a (soft) shadow mapping project and as a first step I want to visualize the depth map on screen. I got render to texture working with FBOs (i.e. COLOR0_ATTACHMENT and DEPTH_ATTACHMENT) and was able to output it on a quad. I then proceeded to render directly to the depth texture and wanted to see just the depth texture on the quad in the fancy grayscale.
However, I hit a snag - the texture when rendered would be completely black. After a lot of searching I found
this ATi article and code example on the web that had a depth map visualization included and noticed that their shader actually set the depth map values to the power of 70.
When I implemented that, I got to see the texture, but also noticed that it is extremely layered. (The image displays the camera's depth buffer as a texture on the quad.)
This is how I create my depth texture:
// Create depth texture
GLuint depth2D;
glGenTextures(1, &depth2D);
glBindTexture(GL_TEXTURE_2D, depth2D);
glTexImage2D(GL_TEXTURE_2D, 0, GL_DEPTH_COMPONENT32_ARB, m_uiRTTWidth,
m_uiRTTHeight, 0, GL_DEPTH_COMPONENT, GL_FLOAT, 0);
glBindTexture(GL_TEXTURE_2D, depth2D);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_S, GL_CLAMP_TO_EDGE);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_WRAP_T, GL_CLAMP_TO_EDGE);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_LINEAR);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_LINEAR);
//GLuint fboDepth;
glGenFramebuffersEXT(1, &m_uiRenderToTextureId);
glBindFramebufferEXT(GL_FRAMEBUFFER_EXT, m_uiRenderToTextureId);
// Attach depth texture
glFramebufferTexture2DEXT(GL_FRAMEBUFFER_EXT, GL_DEPTH_ATTACHMENT_EXT,
GL_TEXTURE_2D, depth2D, 0);
// There are no color buffers attached
glDrawBuffer(GL_NONE);
glReadBuffer(GL_NONE);
// get the status of the frame buffer object
GLenum status = glCheckFramebufferStatusEXT(GL_FRAMEBUFFER_EXT);
// unbind the frame buffer object for now
glBindFramebufferEXT(GL_FRAMEBUFFER_EXT, 0);
assert(status==GL_FRAMEBUFFER_COMPLETE_EXT);
After I've done this, I bind the FBO, render the scene, unbind the FBO and render the scene again, this time with quad textured with the depth2D texture.
// render init
[..]
// bind the frame buffer object
glBindFramebufferEXT(GL_FRAMEBUFFER_EXT, m_uiRenderToTextureId);
// save the view port and set it to the size of the texture
glPushAttrib(GL_VIEWPORT_BIT);
glViewport(0, 0, m_uiRTTWidth, m_uiRTTHeight);
glClearColor(0.0f, 0.0f, 0.0f, 1.0f);
glClear(GL_DEPTH_BUFFER_BIT);
glLoadIdentity();
// render the scene
[..]
// restore old view port and set rendering back to default frame buffer
glPopAttrib();
glBindFramebufferEXT(GL_FRAMEBUFFER_EXT, 0);
// render the scene again
[..]
glEnable(GL_TEXTURE_2D);
glBindTexture(GL_TEXTURE_2D, depth2D);
My question is:
- Is this just a visualization/rendering issue and my depth map is correct, or are these the real depth map values and I have a massive problem?
- If those are the real depth values, what am I doing wrong?
I am running this on a Dell XPS M1710 notebook with nVidia GeForce Go 7950 GTX graphics card and Dell-proprietary drivers that cannot be easily updated (or are always way-behind the actual ForceWare drivers) and OpenGL 1.3. The color and depth buffer depths are both 32 bits.
The interesting part is that said ATi example showcased the same "depth layering". I found out that changing the power exponent in the shader only makes the grayscale layers lighter or darker, it does not change the size of the layers.
Any help on this will be appreciated, since I am afraid to go forward with the shadow mapping if I don't know if the actual depth map is correct.
Thanks,