Search code examples
c++openglframebufferpost-processing

Framebuffer cutting out section of models when moved

I am having an issue with implementing a frame buffer for post processing. So far I can draw everything to a textured quad and a apply a shader to change the screen, so in that respect it works.

My issue is when I move my models or camera, the model start to render incorrectly and parts of them get "cut off". See the following screen shots as it is hard to explain.

Image 1Image 2

To avoid putting pages and pages of code, all the rendering works fine without the Frame Buffer and every looks fine if I don't move the camera or models.

I set up the frame buffer like this.

//Set up FBO
glGenFramebuffers(1, &m_FrameBuffer);
glBindFramebuffer(GL_FRAMEBUFFER, m_FrameBuffer);

glActiveTexture(GL_TEXTURE2);
glGenTextures(1, &m_TexColorBuffer);
glBindTexture(GL_TEXTURE_2D, m_TexColorBuffer);


glTexImage2D(
    GL_TEXTURE_2D, 0, GL_RGB, 1280, 720, 0, GL_RGB, GL_UNSIGNED_BYTE, 0
);

glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST);
glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST);

glFramebufferTexture2D(
    GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, GL_TEXTURE_2D, m_TexColorBuffer, 0
);

glGenRenderbuffers(1, &m_RBODepthStencil);
glBindRenderbuffer(GL_RENDERBUFFER, m_RBODepthStencil);
glRenderbufferStorage(GL_RENDERBUFFER, GL_DEPTH_COMPONENT, 1280, 720);

glFramebufferRenderbuffer(
    GL_FRAMEBUFFER, GL_DEPTH_ATTACHMENT, GL_RENDERBUFFER, m_RBODepthStencil
);

glFramebufferTexture(GL_FRAMEBUFFER, GL_COLOR_ATTACHMENT0, m_TexColorBuffer, 0);

GLenum DrawBuffers[1] = { GL_COLOR_ATTACHMENT0 };
glDrawBuffers(1, DrawBuffers); // "1" is the size of DrawBuffers

if (glCheckFramebufferStatus(GL_FRAMEBUFFER) != GL_FRAMEBUFFER_COMPLETE)
    std::cout << "Frame Buffer: Contructor: Issue completing frame buffer" << std::endl;


//Set Up Shader
m_ScreenShader = new Shader("../Assets/Shaders/screen.vert", "../Assets/Shaders/screen.frag");

//Setup Quad VAO

glGenBuffers(1, &m_QuadVBO);
glBindBuffer(GL_ARRAY_BUFFER, m_QuadVBO);
glBufferData(GL_ARRAY_BUFFER, 24 * sizeof(float), points, GL_STATIC_DRAW);

glGenVertexArrays(1, &m_QuadVAO);
glBindVertexArray(m_QuadVAO);
//glEnableVertexAttribArray(0);
glBindBuffer(GL_ARRAY_BUFFER, m_QuadVBO);
//glVertexAttribPointer(0, 3, GL_FLOAT, GL_FALSE, 0, NULL);

GLint posAttrib = glGetAttribLocation(m_ScreenShader->getID(), "position");
glEnableVertexAttribArray(posAttrib);
glVertexAttribPointer(posAttrib, 2, GL_FLOAT, GL_FALSE, 4 * sizeof(GLfloat), 0);

GLint texAttrib = glGetAttribLocation(m_ScreenShader->getID(), "texcoord");
glEnableVertexAttribArray(texAttrib);
glVertexAttribPointer(texAttrib, 2, GL_FLOAT, GL_FALSE, 4 * sizeof(GLfloat), (void*)(2 * sizeof(GLfloat)));

I then Bind it prior to drawing the scene (skybox and models) and then draw it like this.

Unbind();
glBindVertexArray(m_QuadVAO);
glDisable(GL_DEPTH_TEST);
m_ScreenShader->enable();
m_ScreenShader->setUniform1f("time", glfwGetTime());
GLint baseImageLoc = glGetUniformLocation(m_ScreenShader->getID(), "texFramebuffer");
glUniform1i(baseImageLoc, 2);
glActiveTexture(GL_TEXTURE2);
glBindTexture(GL_TEXTURE_2D, m_TexColorBuffer);

glDrawArrays(GL_TRIANGLES, 0, 6);

Any help is appreciated!

Solution

  • When rendering into a framebuffer that has a depth buffer and the depth test is enabled, then the depth buffer must be cleared. The depth buffer must be cleared after binding the framebuffer and before drawing to the framebuffer: 

    gllBindFramebuffer( GL_FRAMEBUFFER, m_FrameBuffer );
glClear( GL_DEPTH_BUFFER_BIT );

    Note, if you do not clear the deep buffer, it retains its content. This causes that the depth test may fail at positions, where the geometry has been in the past. It follows that parts are missing in the geometry, as in the image in the question.