Raw RGB image in memory - how to create a bitmap and palette I can use?

I suspect this is a trivial problem to solve but it's driving me nuts.


I have functioning code that uses LoadBitmapFromBMPFile() which I lifted from Microsoft's web site to load a bitmap and display it so:

Code:

---

HBITMAP hOldBitmap, hBitmap;
HPALETTE hOldPalette, hPalette;
HDC hDC, hMemDC;
BITMAP bm;

[...]

LoadBitmapFromBMPFile( L"somefile.bmp", &hBitmap, &hPalette );

GetObject( hBitmap, sizeof(BITMAP), &bm );
hMemDC = CreateCompatibleDC( hDC );
hOldBitmap = (HBITMAP)SelectObject( hMemDC, hBitmap );
hOldPalette = SelectPalette( hDC, hPalette, FALSE );
RealizePalette( hDC );

BitBlt( hDC, x, y, bm.bmWidth, bm.bmHeight, hMemDC, 0, 0, SRCCOPY );

---

No problem.


At a later point, I have a raw image in memory. RGB format, 8 bits per pixel. I want to BitBlt() this onto the display, same as above.

I have functioning code that takes that image in memory, writes it to a file in BMP format, and then calls LoadBitmapFromBMPFile() as above ... but this is silly. I ought to be able to skip the file writing and reading.

LoadBitmapFromBMPFile() starts off so:

Code:

---

BOOL LoadBitmapFromBMPFile( LPTSTR szFileName, HBITMAP *phBitmap,
  HPALETTE *phPalette )
  {

  BITMAP  bm;

  *phBitmap = NULL;
  *phPalette = NULL;

  // Use LoadImage() to get the image loaded into a DIBSection
  *phBitmap = (HBITMAP)LoadImage( NULL, szFileName, IMAGE_BITMAP, 0, 0,
              LR_CREATEDIBSECTION | LR_DEFAULTSIZE | LR_LOADFROMFILE );

---

Surely there's some analogue to LoadImage() that, instead of taking a filename as a parameter, can accept a pointer to the already-in-memory RGB image data?

If there is, I haven't found it.



Google has suggested I use CreateDIBSection(). This represents my non-successful attempt:

Code:

---

unsigned char *display_image_data;  // image data in 8 bpp RGB format

[...]

BITMAPINFO pbmi;

pbmi.bmiheader.biSize = sizeof(BITMAPINFO);
pbmi.bmiheader.biWidth = DISPLAY_X;  // img width in pixels
pbmi.bmiheader.biHeight = DISPLAY_Y;  // img height in pixels
pbmi.bmiheader.biPlanes = 1;
pbmi.bmiheader.biBitCount = 24;  // also tried 8 here
pbmi.bmiheader.biCompression = BI_RGB;
pbmi.bmiheader.biSizeImage = 0;  // docs say 0 if BI_RGB is used
pbmi.bmiheader.biXPelPerMeter = 0;  // no idea what to do with this
pbmi.bmiheader.biYPelPerMeter = 0;  // or this
pbmi.bmiheader.biClrUsed = 0;
pbmi.bmiheader.biClrImportant = 0;
// pbmi.bmiColors    shouldn't need for RGB

hBitmap = CreateDIBSection( hDC, &pbmi, DIB_RGB_COLORS, (void **)display_image_data, NULL, 0 );

GetObject( hBitmap, sizeof(BITMAP), &bm );
hMemDC = CreateCompatibleDC( hDC );
hOldBitmap = (HBITMAP)SelectObject( hMemDC, hBitmap );
hOldPalette = SelectPalette( hDC, hPalette, FALSE );
RealizePalette( hDC );

BitBlt( hDC, x, y, bm.bmWidth, bm.bmHeight, hMemDC, 0, 0, SRCCOPY );

---

And what I get is a big black rectangle painted to the screen.

I would appreciate some tips here ... thank you.

pbmi.bmiheader.biSize = sizeof(BITMAPINFO);

shouldn't this be
pbmi.bmiheader.biSize = sizeof(BITMAPINFOHEADER);
http://cboard.cprogramming.com/windo...ng-bitmap.html

Do set the struct to zero (ZeroMemory(&pbmi, sizeof BITMAPINFO) so any members you do not fill will not hold rubbish?

Have you tried SetDiBits() ?
http://cboard.cprogramming.com/windo...p-display.html

Does CreateDIBSection() return an error?

After much wailing and gnashing of teeth, I got it to work. Not sure if this is the "correct" way to do it, but it does work. I used the CreateDIBitmap() function instead.

Code:

---

  BITMAPINFOHEADER BitmapInfoHeader;    BitmapInfoHeader.biSize = sizeof(BITMAPINFOHEADER);  BitmapInfoHeader.biWidth = DISPLAY_X;  BitmapInfoHeader.biHeight = -DISPLAY_Y;  BitmapInfoHeader.biPlanes = 1;  BitmapInfoHeader.biBitCount = 24;  BitmapInfoHeader.biCompression = BI_RGB;  BitmapInfoHeader.biSizeImage = 0;  BitmapInfoHeader.biXPelsPerMeter = 0;  BitmapInfoHeader.biYPelsPerMeter = 0;  BitmapInfoHeader.biClrUsed = 0;  BitmapInfoHeader.biClrImportant = 0;    *phBitmap = CreateDIBitmap( g_hDC, &BitmapInfoHeader, CBM_INIT, (LPVOID *)display_image_data, (LPBITMAPINFO)&BitmapInfoHeader, DIB_RGB_COLORS );

---

I had to flip the sign on the .biHeight member to tell it I wanted 0,0 at the top left rather than the bottom left. And another thing that tripped me up is that Windows wants the data chunk in BGR order, not RGB ... took a few minutes to figure out why the red & blue were flipped on the image. Argh. Why does Microsoft want BGR ordered data?

Code:

---

for( k = 0; k < display_image_length/3; k ++ ) {  temp = display_image_data[k*3];  display_image_data[k*3] = display_image_data[k*3+2];  display_image_data[k*3+2] = temp; }

---

From memory, is over 10 years since I did BMP capture for a RAT app...

Only 24 bit images require the RGB reversed, based on the original OS/R implementation.