OpenGL 2.0 Flip Pixel Data

[N36fSL3A]

Alright, so I'm using OpenGL along with SDL for input and window handling. I'm also using SDL_image to load PNGs into my program.

It's all good, but one problem I have is my PNGs load upside down (Vertically flipped). I've looked all around to fix this, but all I see are fixes for OpenGL ES.

So here's my code:

Texture::Texture(char* location)
{
	// We use SDL_image's function to load the image
	SDL_Surface* img = IMG_Load(location);
	if(img == NULL) {std::cout << "Failed to load " << location << " - " << IMG_GetError() << "\n"; std::cin.get(); exit(4);}

	w = img->w; // We take the width and height from the SDL_Surface*
	h = img->h;

	glGenTextures(1, &texture); // Generate a texture
	glBindTexture(GL_TEXTURE_2D, texture); // Bind that texture temporarily

	int Mode = GL_RGB; // Set the mode

	if(img->format->BytesPerPixel == 4) // If the surface contains an alpha value, tell openGL
	{
		Mode = GL_RGBA;
	}

	glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST); // Set how the texture looks
	glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST); // (Both should be linear)

	// Create the texture. We get the offsets from teh image, then we use it with the image's
	// pixel data to create it.
	glTexImage2D(GL_TEXTURE_2D, 0, Mode, img->w, img->h, 0, Mode, GL_UNSIGNED_BYTE, img->pixels);

	// Unbind the texture
	glBindTexture(GL_TEXTURE_2D, NULL);

	SDL_FreeSurface(img); // Free the SDL_Surface, as we no longer need it :)
}

This loads the PNG, then it converts the SDL_Surface* to an OpenGL texture. What I want to do is change the pixel data so it gets flipped right side up, but I don't know where to start.

Thanks in advance.

[cire]

The simplest solution would be to load your image in an image editing program, flip it and save it.

Or, invert the y/v coordinate in your shader.

Manually flipping the image data before feeding it to glTexImage2D is also an option.

[N36fSL3A]

cire wrote:

Manually flipping the image data before feeding it to glTexImage2D is also an option.

That's what I was looking to do, the problem is I don't know how to do it.

[cire]

#include <iostream>

const unsigned rows =  6;
const unsigned cols = 12;

const char image [rows][cols] =
{
    { "     #     " },
    { "    # #    " },
    { "   #   #   " },
    { "  #     #  " },
    { " #       # " },
    { "###########" }
};

int main()
{
    char flippedImage[rows][cols];

    for (unsigned i = 0; i < rows; ++i)
        for (unsigned j = 0; j < cols; ++j)
            flippedImage[i][j] = image[rows-i-1][j];

    std::cout << "Original:\n";
    for (unsigned i = 0; i < rows; ++i)
        std::cout << image[i] << '\n';

    std::cout << "\nFlipped:\n";
    for (unsigned i = 0; i < rows; ++i)
        std::cout << flippedImage[i] << '\n';
}

http://ideone.com/sMzBe9

Except you're working with pixels instead of chars.

[N36fSL3A]

But I have a void pointer of pixels. How would I do that?

[qZanity]

Try setting the mode to GL_BGR

Edit: oops scratch that thought you meant colours are inverted.

Yes do what cire says, I would create a struct, then copy the pointer contents to the struct... And invert much cleaner that way.

[cire]

But I have a void pointer of pixels. How would I do that?

You have the address of the pixels. You know functions and algorithms that work with pointers, right?

I don't have SDL installed, but from looking at the documentation, something like the following (obviously untested) code:

Texture::Texture(char* location)
{
    // We use SDL_image's function to load the image
    SDL_Surface* img = IMG_Load(location);
    if (img == NULL) { std::cout << "Failed to load " << location << " - " << IMG_GetError() << "\n"; std::cin.get(); exit(4); }

    w = img->w; // We take the width and height from the SDL_Surface*
    h = img->h;

    typedef uint8_t byte_type;

    const unsigned bytesPerRow = img->pitch;
    const byte_type* originalRaw = static_cast<byte_type*>(img->pixels);
    byte_type* flippedRaw = new byte_type[bytesPerRow * img->h];

    for (unsigned i = 0; i < img->h; ++i)
    {
        const byte_type* srcBeg = originalRaw + (bytesPerRow *(img->h - i - 1));
        const byte_type* srcEnd = srcBeg + bytesPerRow;

        std::copy(srcBeg, srcEnd, flippedRaw + bytesPerRow * i);
    }

    glGenTextures(1, &texture); // Generate a texture
    glBindTexture(GL_TEXTURE_2D, texture); // Bind that texture temporarily

    int Mode = GL_RGB; // Set the mode

    if (img->format->BytesPerPixel == 4) // If the surface contains an alpha value, tell openGL
    {
        Mode = GL_RGBA;
    }

    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MIN_FILTER, GL_NEAREST); // Set how the texture looks
    glTexParameteri(GL_TEXTURE_2D, GL_TEXTURE_MAG_FILTER, GL_NEAREST); // (Both should be linear)

    // Create the texture. We get the offsets from teh image, then we use it with the image's
    // pixel data to create it.
    glTexImage2D(GL_TEXTURE_2D, 0, Mode, img->w, img->h, 0, Mode, GL_UNSIGNED_BYTE, flippedRaw);    // *

    // Unbind the texture
    glBindTexture(GL_TEXTURE_2D, NULL);

    delete [] flippedRaw;                         // *
    SDL_FreeSurface(img);
}

[edit: spelling]

[Lowest0ne]

There is absolutly no reason to flip your image.

You dont show the code you use to actually render it to the screen.

I guess you follow this ?
http://www.sdltutorials.com/sdl-tip-sdl-surface-to-opengl-texture

Make sure that your texture coordinates are in the correct order ( glTexCoord2f() ).

I can post tomorrow the code I used.

[Lowest0ne]

Yep, my code is a little different than the code at that site:

glBegin(GL_QUADS);
    glTexCoord2f(0, 0); glVertex3f(rect.topLeft.x     , rect.topLeft.y    , 0); // 1
    glTexCoord2f(0, 1); glVertex3f(rect.topLeft.x     , rect.bottomRight.y, 0); // 4
    glTexCoord2f(1, 1); glVertex3f(rect.bottomRight.x , rect.bottomRight.y, 0); // 3
    glTexCoord2f(1, 0); glVertex3f(rect.bottomRight.x , rect.topLeft.y    , 0); // 2
glEnd();

Judging by my comments, I had trouble too :)

[cire]

Yep, my code is a little different than the code at that site:

You inverted the texture v coordinate. I mentioned that as an option in my first post as something that could be done in the shader although I suppose it can just as easily be done in the coordinates you feed to the shader. (You were also using deprecated functions.)

[Lowest0ne]

(You were also using deprecated functions.)

For sure.

That was for drawing a texture in 2d space ( a UI element ), probably should have been using glDrawPixels() anyway.

... Such is the code when you try to do everything yourself.

[N36fSL3A]

EDIT: Thanks cire, works perfectly. :D

This is off topic, but how do I save/load a float to/from a binary file? It has to work on both 64 and 32 bit machines.

[N36fSL3A]

bump

[cire]

Why are you bumping a thread to get an answer to a question that has nothing to do with the thread? Google "C++ float serialization."

http://stackoverflow.com/questions/4733147/portability-of-binary-serialization-of-double-float-type-in-c

And make a new topic if you have further questions.