双抛物面阴影映射

Question

我使用的是OpenGL 4.0，GLSL 4.0，不是GLEW或类似的，不是GLU或类似的，不是GLUT或类似的。也使用OpenCL或CUDA，但它们不涉及这种情况。

我一直试图解决我的问题几周没有成功，现在我希望有双抛物面阴影映射经验的人可以给我建议。让我们直接解决问题并检查一些图片（图片1）：

图1包含一些我需要解释的彩色箭头。红色箭头显示我们应该看到的唯一正确的阴影。其他箭头表示阴影错误：黄色箭头表示由细分引起的斑点，也许是蓝色箭头，但它们的位置在前/后半球的边界上。并且绿色箭头指向不应存在的锯齿图案（我们最近看到的例子中没有）。现在重要的是要注意上面的图片是使用以下代码行（代码1）计算的：

"swap.z=-sign(swap.z)*swap.z;\n" //mostly right results in the main project, but wrong way

这个代码行位于GLSL着色器程序中，它是我在主项目中尝试没有成功的四个候选者之一，从中拍摄照片。但是，正如我们将看到的，代码4确实在单独的测试程序中工作。代码1实际上是完全错误的DPSM方式，但它是我的主项目获得阴影的唯一方式。接下来我们看相同的场景计算有点不同，但仍然错误的代码行（图片2和代码2）：

"swap.z=sign(swap.z)*swap.z;\n" //almost dark picture, wrong way

我们再次看同一个场景，但现在我们使用完全不同的正统代码行（图3和代码3）：

"swap.z=-(teDistance-n)/(f-n);\n" //lightning is mainly working but no shadows, should be the right way

最后，我们看一下代码行计算的场景，它在我们最近看到的例子中（几乎）完美地工作（图4和代码4）：

"swap.z=(teDistance-n)/(f-n);\n" //almost dark picture, doesn't work in the main project, but works in the test program, right way

如果有人怀疑上面图片中看到的文物是由于“阴影痤疮”现象造成的，那么不，我认为它们不是。以下是通过设置SHADOW_EPSILON = 0.000005f并关闭模糊（图5）故意制作阴影痤疮图案的图片：

在这一点上，我需要说，我在两台独立的Windows 7.1笔记本电脑上运行程序，一台配备nVIDIA GeForce GT 525M，另一台配备AMD Radeon R6。结果是一样的。编译器是Visual Studio 2010.在我看来，这是一个纯粹的OpenGL相关问题。

为了解决这个问题，我写了一个单独的小测试程序，最后得到了阴影贴图。据我所知，测试程序的工作方式与制作图片1-5的程序非常相似，但没有优化，并且许多矩阵乘法已经从主机移动到着色器。测试程序源的相关部分如下。着色器优先：

    static const char *vertex1=
"#version 400 core\n"

"layout (location=1) in vec3 vertexLocation;\n"

"out vec3 vPosition;\n"

"void main() {\n"
    "vPosition=vertexLocation;\n"
"}\0";




static const char *tessIn1=
"#version 400 core\n"

"layout (vertices=3) out;\n"

"in vec3 vPosition[];\n"
"out vec3 tcPosition[];\n"

"void main() {\n"
    "tcPosition[gl_InvocationID]=vPosition[gl_InvocationID];\n"
    "if (gl_InvocationID==0) {\n"

        "gl_TessLevelOuter[0]=max(distance(vPosition[1], vPosition[2]), 1.0);\n"
        "gl_TessLevelOuter[1]=max(distance(vPosition[2], vPosition[0]), 1.0);\n"
        "gl_TessLevelOuter[2]=max(distance(vPosition[0], vPosition[1]), 1.0);\n"
        "gl_TessLevelInner[0]=max(0.33*(gl_TessLevelOuter[0]+gl_TessLevelOuter[1]+gl_TessLevelOuter[2]), 1.0);\n"

    "}\n"
"}\0";


static const char* tessOut1=
"#version 400 core\n"

"layout(triangles, equal_spacing, ccw) in;\n"

"uniform mat4 model;\n"
"uniform mat4 view;\n"
"uniform mat4 lightOrientation;\n"

"in vec3 tcPosition[];\n"
"out float teDistance;\n"
"out float teClip;\n"

"const float n=0.5;\n"
"const float f=20000.0;\n"

"void main() {\n"

    "vec3 accum=vec3(0.0);\n"
    "accum=accum+gl_TessCoord.x*tcPosition[0];\n"
    "accum=accum+gl_TessCoord.y*tcPosition[1];\n"
    "accum=accum+gl_TessCoord.z*tcPosition[2];\n"

// Transform position to the paraboloid's view space
    "vec4 swap=lightOrientation*model*vec4(accum, 1.0);\n"

//store the distance and other variables
    "teDistance=abs(swap.z);\n"
    "teClip=swap.z;\n"

//calculate and set X and Y coordinates
    "swap.xyz=normalize(swap.xyz);\n"
    "if (swap.z<=0.0) {\n"
        "swap.xy=swap.xy/(1.0-swap.z);\n"
    "} else {\n"
        "swap.xy=swap.xy/(1.0+swap.z);\n"
    "}\n"

//calculate and set Z and W coordinates
//  "swap.z=-sign(swap.z)*swap.z;\n"  //Wrong way
//  "swap.z=sign(swap.z)*swap.z;\n"  //Wrong way
//  "swap.z=-(teDistance-n)/(f-n);\n"      //Wrong way
    "swap.z=(teDistance-n)/(f-n);\n"      //Right way
    "swap.w=1.0;\n"

    "gl_Position=swap;\n"
"}\0";


static const char* geometry1=
"#version 400 core\n"

"layout(triangles) in;\n"
"layout(triangle_strip, max_vertices=3) out;\n"

"in float teDistance[];\n"
"in float teClip[];\n"
"out float gDistance;\n"

"void main() {\n"
    "for (int i=0; i<3; i++) {\n"
        "gDistance=teDistance[i];\n"
        "if (teClip[i]<=0.0) {\n"
            "gl_Layer=0;\n"
        "} else {\n"
            "gl_Layer=1;\n"
        "}\n"
        "gl_Position=gl_in[i].gl_Position;\n"
        "EmitVertex();\n"
    "}\n"
    "EndPrimitive();\n"
"}\0";


static const char* fragment1=
"#version 400 core\n"

"in float gDistance;\n"

"out vec2 fragmentVari;\n"

"void main() {\n"

    "fragmentVari=vec2(gDistance, gDistance*gDistance);\n"

"}\0";


const char *vertex2=
"#version 400 core\n"

"layout (location=1) in vec3 vertexPosition;\n"
"layout (location=2) in vec2 vertexTexCoord;\n"
"layout (location=3) in vec3 vertexNormal;\n"

"const float n=0.5;\n"
"const float f=20000.0;\n"

"uniform vec4 colour;\n"
"uniform mat4 model;\n"
"uniform mat4 view;\n"
"uniform mat4 normal;\n"
"uniform mat4 projection;\n"
"uniform mat4 lightOrientation;\n"

"out vec2 texKoord;\n"
"out vec3 pointNormal;\n"
"out vec3 point;\n"
"out vec4 color;\n"
"out vec4 vOriginPoint;\n"

"void main() {\n"
    "texKoord=vertexTexCoord;\n"
    "pointNormal=normalize(vec3(normal*vec4(vertexNormal, 1.0)));\n"
    "point=vec3(model*vec4(vertexPosition, 1.0));\n"
    "color=colour;\n"
    "vOriginPoint=vec4(vertexPosition, 1.0);\n"
    "gl_Position=projection*view*model*vec4(vertexPosition, 1.0);\n"
"}\0";


const char *fragment2=
"#version 400 core\n"

"uniform sampler2DArray tex1;\n"

"uniform vec4 colour;\n"
"uniform mat4 model;\n"
"uniform mat4 view;\n"
"uniform mat4 normal;\n"
"uniform mat4 projection;\n"
"uniform mat4 lightOrientation;\n"

"in vec2 texKoord;\n"
"in vec3 pointNormal;\n"
"in vec3 point;\n"
"in vec4 color;\n"
"in vec4 vOriginPoint;\n"
"out vec4 fragmentColor;\n"

"const float SHADOW_EPSILON = 0.05f;\n"
"const vec3 Ka=vec3(0.05, 0.05, 0.05);\n"  //Ambient reflectivity
"const vec3 Kd=vec3(1.0, 1.0, 1.0);\n"  //Diffuse reflectivity
"const float At=0.4;\n"                 //Light attenuation

"vec3 ads(in vec3 position, in vec3 normal) {\n"
    "vec3 l=vec3(lightOrientation*model*vOriginPoint);\n"
    "vec3 s=normalize(l - position);\n"
    "vec3 intensity=vec3(0.5, 0.5, 0.5)*10.0;\n"
    "float attenuation=1.0/(1.0+At*max(length(l), 1.0));\n"
    "intensity=intensity*attenuation*Kd*abs(dot(s, normal));\n"
    "return intensity;\n"
"}\n"

"float drawShadow() {\n"
    "vec3 texKoord;\n"
    "vec4 textureDepth;\n"

    "vec4 originPoint=vec4(lightOrientation*model*vOriginPoint);\n"
    "float distance=abs(originPoint.z);\n"
    "vec3 normalized=normalize(originPoint.xyz);\n"

    "if (normalized.z<=0.0) {\n"
        "texKoord.xy=normalized.xy/(1.0-normalized.z);\n"
        "texKoord.xy=0.5*texKoord.xy+0.5;\n"
        "texKoord.z=0.0;\n"
        "textureDepth=texture(tex1, texKoord);\n"
    "} else {\n"
        "texKoord.xy=normalized.xy/(1.0+normalized.z);\n"
        "texKoord.xy=0.5*texKoord.xy+0.5;\n"
        "texKoord.z=1.0;\n"
        "textureDepth=texture(tex1, texKoord);\n"
    "}\n"

    "if (textureDepth.x+SHADOW_EPSILON>=distance) {\n"
        "return 1.0;\n"
    "} else {\n"
        "return 0.0;\n"
    "}\n"
"}\n"

"void main() {\n"
    "vec4 lightning=vec4(Ka, 1.0);\n"
    "swap2=swap2*drawShadow();\n"
    "lightning=lightning+swap2;\n"
    "fragmentColor=color*lightning;\n"
"}\0";


const char *vertexLight=
"#version 400 core\n"

"layout (location=1) in vec3 vertexPosition;\n"

"uniform mat4 view;\n"
"uniform mat4 projection;\n"
"uniform mat4 lightOrientation;\n"

"void main() {\n"
    "gl_Position=projection*view*lightOrientation*vec4(vertexPosition, 1.0);\n"
"}\0";


const char *fragmentLight=
"#version 400 core\n"

"out vec4 fragmentColor;\n"

"void main() {\n"
    "fragmentColor=vec4(1.0, 1.0, 1.0, 1.0);\n"
"}\0";

这是显示功能：

void TForm1::display()
{
    GLuint loc1, loc2, loc3, loc4, loc5, loc6;
    float swap[16];
    float normal[16]={0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 0.0, 1.0};

//first we render a shadow map
    {
        glUseProgram(shaderT1);
        glBindFramebuffer(GL_FRAMEBUFFER, frameBuffer[0]);
        glClearColor(20000.0f, 0.0f, 0.0f, 0.0f);
        glDepthMask(GL_TRUE);
        glDepthRange(0, 1);
        glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
        glViewport(0, 0, textureDim.x, textureDim.y);
        glPatchParameteri(GL_PATCH_VERTICES, 3);
        loc1=glGetUniformLocation(shaderT1, "model\0");
        loc2=glGetUniformLocation(shaderT1, "lightOrientation\0");
        loc3=glGetUniformLocation(shaderT1, "view\0");
        swap[0]=1.0; swap[1]=0.0; swap[2]=0.0; swap[3]=0.0;
        swap[4]=0.0; swap[5]=1.0; swap[6]=0.0; swap[7]=0.0;
        swap[8]=0.0; swap[9]=0.0; swap[10]=1.0; swap[11]=0.0;
        swap[12]=-lightMatrix[12]; swap[13]=-lightMatrix[13]; swap[14]=-lightMatrix[14]; swap[15]=1.0;
        glUniformMatrix4fv(loc1, 1, GL_FALSE, triangleMatrix);
        glUniformMatrix4fv(loc2, 1, GL_FALSE, swap);
        glUniformMatrix4fv(loc3, 1, GL_FALSE, view);
        glBindVertexArray(VAO[1]);
        glDrawArrays(GL_PATCHES, 0, 3);
        glUniformMatrix4fv(loc1, 1, GL_FALSE, identity);
        glUniformMatrix4fv(loc2, 1, GL_FALSE, swap);
        glUniformMatrix4fv(loc3, 1, GL_FALSE, view);
        glBindVertexArray(VAO[0]);
        glDrawArrays(GL_PATCHES, 0, 6);
    }

//then we render the world and make use of that rendered shadow map
    {
        glUseProgram(shaderT2);
        glBindFramebuffer(GL_FRAMEBUFFER, 0);
        glDepthMask(GL_TRUE);
        glDepthRange(0, 1);
        glClearColor(0.0f, 1.0f, 0.0f, 0.0f);
        glClear(GL_COLOR_BUFFER_BIT | GL_DEPTH_BUFFER_BIT);
        glUniform1i(glGetUniformLocation(shaderT2, "tex1"), 1);
        glActiveTexture(GL_TEXTURE0+1);
        glBindTexture(GL_TEXTURE_2D_ARRAY, texID[0]);
        glViewport(0, 0, 512, 512);
        loc1=glGetUniformLocation(shaderT2, "model\0");
        loc2=glGetUniformLocation(shaderT2, "view\0");
        loc3=glGetUniformLocation(shaderT2, "normal\0");
        loc4=glGetUniformLocation(shaderT2, "colour\0");
        loc5=glGetUniformLocation(shaderT2, "projection\0");
        loc6=glGetUniformLocation(shaderT2, "lightOrientation\0");

//render a rectangle where the shadow is drawn onto
        glUniformMatrix4fv(loc1, 1, GL_FALSE, identity);
        glUniformMatrix4fv(loc2, 1, GL_FALSE, view);
        matrixMultiply4D(swap, view, identity);
        inverseMatrix4D(swap, swap);
        transpose4D(normal, swap);
        glUniformMatrix4fv(loc3, 1, GL_FALSE, normal);
        glUniform4fv(loc4, 1, red);
        glUniformMatrix4fv(loc5, 1, GL_FALSE, projection);
        swap[0]=1.0; swap[1]=0.0; swap[2]=0.0; swap[3]=0.0;
        swap[4]=0.0; swap[5]=1.0; swap[6]=0.0; swap[7]=0.0;
        swap[8]=0.0; swap[9]=0.0; swap[10]=1.0; swap[11]=0.0;
        swap[12]=-lightMatrix[12]; swap[13]=-lightMatrix[13]; swap[14]=-lightMatrix[14]; swap[15]=1.0;
        glUniformMatrix4fv(loc6, 1, GL_FALSE, swap);
        glBindVertexArray(VAO[0]);
        glDrawArrays(GL_TRIANGLES, 0, 6);

//render the triangle which makes a shadow
        glUniformMatrix4fv(loc1, 1, GL_FALSE, triangleMatrix);
        glUniformMatrix4fv(loc2, 1, GL_FALSE, view);
        matrixMultiply4D(swap, view, triangleMatrix);
        inverseMatrix4D(swap, swap);
         transpose4D(normal, swap);
        glUniformMatrix4fv(loc3, 1, GL_FALSE, normal);
        glUniform4fv(loc4, 1, yellow);
        glUniformMatrix4fv(loc5, 1, GL_FALSE, projection);
        swap[0]=1.0; swap[1]=0.0; swap[2]=0.0; swap[3]=0.0;
        swap[4]=0.0; swap[5]=1.0; swap[6]=0.0; swap[7]=0.0;
        swap[8]=0.0; swap[9]=0.0; swap[10]=1.0; swap[11]=0.0;
        swap[12]=-lightMatrix[12]; swap[13]=-lightMatrix[13]; swap[14]=-lightMatrix[14]; swap[15]=1.0;
        glUniformMatrix4fv(loc6, 1, GL_FALSE, swap);
        glBindVertexArray(VAO[1]);
        glDrawArrays(GL_TRIANGLES, 0, 3);
    }

//finally render a white triangle which represents a location of the light
    {
        glUseProgram(shaderT3);
        glBindFramebuffer(GL_FRAMEBUFFER, 0);
        glDepthMask(GL_TRUE);
        glDepthRange(0, 1);
        glViewport(0, 0, 512, 512);
        loc1=glGetUniformLocation(shaderT3, "view\0");
        loc2=glGetUniformLocation(shaderT3, "projection\0");
        loc3=glGetUniformLocation(shaderT3, "lightOrientation\0");
        glUniformMatrix4fv(loc1, 1, GL_FALSE, view);
        glUniformMatrix4fv(loc2, 1, GL_FALSE, projection);
        glUniformMatrix4fv(loc3, 1, GL_FALSE, lightMatrix);
        glBindVertexArray(VAO[2]);
         glDrawArrays(GL_TRIANGLES, 0, 3);
    }

    glFinish();

//rotate a light on it's orbit
    matrixMultiply4D(lightMatrix, rotationMatrix, lightMatrix);

}

我拍了四部电影代表上面的测试程序和代码行1-4。下面是用代码1完成的电影1：

电影2代码2：

电影3代码3：

电影4代码4：

你可以看到Movie 4是唯一一个阴影贴图工作的地方（两个半球之间的边界区域没有阴影，但我可以忍受。但是，如果有人知道如何修复它，我会很高兴如果你告诉我）。但它不适用于主项目！我希望你能给我一些关于可能出错的建议，我应该检查或给你你的双抛物面阴影映射代码样本......