Creating BMP File

I've been working for a while on image processing and I've noticed weird things. I'm reading a BMP file, using simple methods like ReadFile and stuff, and using Microsoft's BMP structures. Here is the code:

 ReadFile(_bmpFile,&bmpfh,sizeof(bfh),&data,NULL);
 ReadFile(_bmpFile, &bmpih, sizeof(bih), &data, NULL);
 imagesize = bih.biWidth*bih.biHeight;
 image = new RGBQUAD[imagesize];
 ReadFile(_bmpFile,image, imagesize*sizeof(RGBQUAD),&written,NULL);

That is how I read the file and then I'm turning it into gray scale using a simple for-loop.

  for (int i = 0; i < imagesize; i++)
  {
       RED = image[i].rgbRed;
       GREEN = image[i].rgbGreen;
       BLUE = image[i].rgbBlue;
       avg = (RED + GREEN + BLUE ) / 3;
       image[i].rgbRed = avg;
       image[i].rgbGreen = avg;
       image[i].rgbBlue = avg;
  }

Now when I write the file using this code:

#pragma pack(push, 1)
WriteFile(_bmpFile, &bmpfh, sizeof(bfh), &data, NULL);
WriteFile(_bmpFile, &bmpih, sizeof(bih), &data, NULL);
WriteFile(_bmpFile, image, image开发者_Python百科size*sizeof(RGBQUAD), &written, NULL);
#pragma pack(pop)

The file is getting much bigger(30MB -> 40MB).

The reason it happens is because I'm using RGBQUAD instead RGBTRIPLE, but if i'm using RGBTRIPLE I have a problem converting small pictures into gray scale - can't open the picture after creating it(says it's not in the right structure).

Also the file size is missing one byte, (1174kb and after 1173kb)

Has anybody seen this before (it only occurs with small pictures)?

In a BMP file, every scan line has to be padded out so the next scan line starts on a 32-bit boundary. If you do 32 bits per pixel, that happens automatically, but if you use 24 bits per pixel, you'll need to add code to do it explicitly.

You are ignoring stride (Jerry's comment) and the pixel format of the bitmap. Which is 24bpp judging by the file size increase, you are writing it as though it is 32bpp. Your grayscale conversion is wrong, the human eye isn't equally sensitive to red, green and blue.

Consider using GDI+, you #include <gdiplus.h> in your code to use the Bitmap class. Its LockBits() method gives you access to the bitmap bits. The ColorMatrixEffect class lets you apply a color transformation in a single operation. Check this answer for the color matrix you need to get a grayscale image. The MSDN docs start here.

Each horizontal row in a BMP must be a multiple of 4 bytes long.

If the pixel data does not take up a multiple of 4 bytes, then 0x00 bytes are added at the end of the row. For a 24-bpp image, the number of bytes per row is (imageWidth*3 + 3) & ~3. The number of padding bytes is ((imageWidth*3 + 3) & ~3) - (imageWidth*3).

This was answered by immibis.

I would like to add that the size of array is ((imageWidth*3 + 3) & ~3)*imageHeight. I hope this helps