Problems using GLTexImage3D correctly

问题

This is how I give the Bitmaps to OpenGL. (C#)

public static int Generate3DTexture( string[] names ) // file paths
{
    //basically merging the images into one vertical column of images
    MagickImageCollection allimages = new MagickImageCollection();
    foreach (string eachname in names)
        allimages.Add(new MagickImage(eachname));
    MagickImage template = new MagickImage(MagickColor.FromRgba(0, 0, 0, 0), MasterSize.Width, MasterSize.Height * names.Length);
    Point drawpnt = new Point(0, 0);
    foreach (MagickImage each in allimages)
    {
        template.Composite(each, drawpnt.X, drawpnt.Y, CompositeOperator.Overlay);
        drawpnt.Y += each.Height;
    }
    Bitmap merged = template.ToBitmap();
    //saving the bitmap here is confirmed to stack them vertically
    BitmapData thedata = merged.LockBits(new Rectangle(new Point(0, 0), merged.Size), ImageLockMode.ReadOnly, WfPixelFormat.Format32bppArgb);
    int texId = GL.GenTexture();
    GL.BindTexture(TextureTarget.Texture3D, texId);
    GL.TexImage3D(TextureTarget.Texture3D, 0, PixelInternalFormat.Rgba, MasterSize.Width, MasterSize.Height, names.Length, 0, GlPixelFormat.Bgra, PixelType.UnsignedByte, thedata.Scan0);
    GL.GenerateMipmap(GenerateMipmapTarget.Texture3D);
    GL.BindTexture(TextureTarget.Texture3D, 0);
    merged.UnlockBits(thedata);
    return texId;
}

I found a description for how the data should be organized here, which I interpreted to mean the images should be arranged vertically, and possibly flipped or something.

So for my test I used 2 images, on the fragment shader I apply the texture like so:

void main()
{
    outputColor =  texture3D(ThreeDTexture,vec3(texcoord.x, 1.0 - texcoord.y, 0));
}

The result is the two images merged in roughly equal proportions. That's what I would expect if I made the vec3.z parameter 0.5. So I tried it with three images. It gives a 50-50 combination of the first and last images supplied (?!?!). This is it's behavior without regard to the value I give for vec3.z.

I expected that parameter to be the Z axis for the images supplied. Or if I used whole numbers for it to be an index for the array of images supplied.

EDIT: It is the index, but projected on the range 0 through 1, like the x,y coordinates are.

Where did I go wrong?

What I'm giving to OpenGL (in order):

The result:

EDIT: As Rabbid76 explains, the solution was to insert this after the call to GL.TexImage3D:

    GL.TexParameter(TextureTarget.Texture3D, TextureParameterName.TextureWrapS, (int)TextureParameterName.ClampToEdge);
    GL.TexParameter(TextureTarget.Texture3D, TextureParameterName.TextureWrapT, (int)TextureParameterName.ClampToEdge);
    GL.TexParameter(TextureTarget.Texture3D, TextureParameterName.TextureWrapR, (int)TextureParameterName.ClampToEdge);

回答1:

The texture wrap parameters GL_TEXTURE_WRAP_S, GL_TEXTURE_WRAP_T and GL_TEXTURE_WRAP_R are by default GL_REPEAT.
The default parameters for the minifying function (GL_TEXTURE_MIN_FILTER) and magnification function are (GL_TEXTURE_MAG_FILTER) are GL_NEAREST_MIPMAP_LINEAR respectively GL_LINEAR.
See glTexParameter.

The combination of "REPEAT" and "LINEAR" causes, that the first voxel is mixed (interpolated) with the last voxel of a row, column, or depth-layer, if the the texture coordinate parameters, which is passed to the lookup function, is 0.0.

If you would use the wrap parameter GL_CLAMP_TO_EDGE, then the first and the last voxel won't become mixed, because the texture coordinate is clamped.

Note the texture coordinate of the first voxel (or texel) is 1/(2*N) and the coordinate of the last voxel is 1 - 1/(2*N), where N is the number of voxels in a row, column or layer. Because of that the coordinate 0.0, is exactly in the middle of the first and the last voxel. GL_REPEAT would clamp the coordinate 0.0 to 1/(2*N).

来源：https://stackoverflow.com/questions/52326761/problems-using-glteximage3d-correctly

标签

opengl

glsl

opentk