Building the Right Environment to Support AI, Machine Learning and Deep Learning
The basic problem of this code was to decrease a color depth of great bitmaps (like 8000 x 8000 pixels) to - for example - 4 bits per pixel and solving nearly (! see below) the original color palette. The first attempt was to use the method as shown in DDBtoDIB. But such great bitmaps produce an error: They do nothing sencefull in functions like GetDIBits. So I handle the bitmap in bands of smaller bitmap data and make direct file reading/saving operations of the bitmaps.
Because of the internal mapping of the color palette to the system palette I got some strange results. Also, by realizing the whole number of really used colors, I ve got another problem with the previous banding of the bitmap, which show me only a small part of the overall used colors. This was the second problem. So I create a color palette by counting and adding the colors of each band to a global color palette of the bitmap.
The result is a class BitmapLib with two (static) methods, ChangeColorDepth() and GetUsedColors().
BOOL CBitmapLib::ChangeColorDepth( CString const & stSrcBitmapFile, CString const & stDstBitmapFile, WORD const wDstBitCount, CProgressCtrl * pProgressCtrl = NULL );
ChangeColorDepth() changes a color depth of a file based bitmap and saves the result into a destination file. The operation is done by increasing or decreasing the bits per pixel through the whole range from 1 to 32 (thus converting also from device dependent into device independent or the way back). The progress control is only used if nescessary and showing the work progress (great bitmaps last a few seconds ...)
int CBitmapLib::GetUsedColors( CString const & stSrcBitmapFile, RGBQUAD * pColorTable, WORD const wMaxColorTableSize, CProgressCtrl * pProgressCtrl = NULL ); int CBitmapLib::GetUsedColors( BITMAPINFO const * pBmi, BYTE const * pBits, UINT const nStartScanLine, UINT const nScanLines, RGBQUAD * pColorTable, WORD const wMaxColorTableSize, WORD & wUsedColors );
A byproduct of the operation ChangeColorDepth() is the method GetUsedColors(), which count the the really used colors and produces a color table of the used colors.
Both methods abort its operation, if the destination color depth is not sufficient enough for the number of used colors in the source bitmap. This means that you always must have a raw idea of the final color table result - you cannot convert any High/True color bitmap with an unlimited size of used colors down to a DIB! Therefore you will need some intelligent algorithm, which must be incorporated within the library.