Question

I have a method to copy the data out of a System.Drawing.Bitmap which looks like this:

var readLock = image.LockBits(new Rectangle(0, 0, image.Width, image.Height), ImageLockMode.ReadOnly, PixelFormat.Format24bppRgb);
byte[] data = new byte[3 * image.Width * image.Height];
if (data.Length != readLock.Stride * readLock.Height)
    throw new InvalidOperationException("Incorrect number of bytes");
Marshal.Copy(readLock.Scan0, data , 0, data.Length);
image.UnlockBits(readLock);

Pretty simple, and it works for most of my images. However for a very small image (14×14) it hits the exception. In the failing case Stride is 44, not 42 (14 * 3) as expected.

The pixel format is Format24bppRgb, so there should be three bytes for every pixel in the image. Where are these extra bytes coming from, and how can I deal with them when processing the image data?

For anyone interested, I’m generating Normal data from a heightmap, so I need to be able to get each pixel and its neighbours accurately).

Answer 1

Every pixel line of Bitmap must be aligned, that’s why stride is not always width * bytes-per-pixel. You should ignore any extra bytes. It means that if you are working with byte arrays with unaligned data, you might not always be able to copy all image data in a single Marshal.Copy() call. Every line of pixels starts at readLock.Scan0 + y * readLock.Stride and contains readLock.Width * bytes-per-pixel meaningful bytes.

Solution:

const int BYTES_PER_PIXEL = 3;
var data = new byte[readLock.Width * readLock.Height * BYTES_PER_PIXEL];
if(readLock.Stride == readLock.Width * BYTES_PER_PIXEL)
{
    Marshal.Copy(readLock.Scan0, data, 0, data.Length);
}
else
{
    for(int y = 0; y < readLock.Height; ++y)
    {
        IntPtr startOfLine = (IntPtr)((long)readLock.Scan0 + (readLock.Stride * y));
        int dataOffset = y * readLock.Width * BYTES_PER_PIXEL;
        Marshal.Copy(startOfLine, data, dataOffset, readLock.Width * BYTES_PER_PIXEL);
    }
}