Converting a bitmap to a byte array

[MrPolite]

this comes from my own l33t hax0r skillz
if you've got a better way, say so

method 1 just copies the bitmap bytes from memory
method 2 actually saves the image and writes any image headers and whatnot using a proper image format, and then converts the written data to a byte array
method 3 saves the objects data using serialization (very useless just serialize to file or something... included for the heck of it)


the l33test way in my opinion is using unsafe code:

Method 1: unsafe code, locking the bitmap and using the Marshal class

// import this 
// using System.Runtime.InteropServices;
private unsafe byte[] BmpToBytes_Unsafe (Bitmap bmp)
{
	BitmapData bData = bmp.LockBits(new Rectangle (new Point(), bmp.Size),
		ImageLockMode.ReadOnly, 
		PixelFormat.Format24bppRgb);
	// number of bytes in the bitmap
	int byteCount = bData.Stride * bmp.Height;
	byte[] bmpBytes = new byte[byteCount];

	// Copy the locked bytes from memory
	Marshal.Copy (bData.Scan0, bmpBytes, 0, byteCount);

	// don't forget to unlock the bitmap!!
	bmp.UnlockBits (bData);

	return bmpBytes;
}

in most cases this works fine, because your bitmap doesn't have an alpha channel. But if it does, and you want to preserve it, then use Format32bppArgb in the fist line when locking the bitmap.
I wish I could think of a way of doing this without locking the bitmap (ie, you could use GCHandle.Alloc() and then call Marshal.Copy() using the created handle, but the problem is I wouldnt know the size of the bitmap without locking it's bits and the Marshal.Copy() function needs to know the size).

reversing this process is a bit ugly, as you need to know the dimensions of the bitmap.

private unsafe Bitmap BytesToBmp (byte[] bmpBytes, Size imageSize)
{
	Bitmap bmp = new Bitmap (imageSize.Width, imageSize.Height);

	BitmapData bData  = bmp.LockBits (new Rectangle (new Point(), bmp.Size),
		ImageLockMode.WriteOnly,
		PixelFormat.Format24bppRgb);

	// Copy the bytes to the bitmap object
	Marshal.Copy (bmpBytes, 0, bData.Scan0, bmpBytes.Length);
	bmp.UnlockBits(bData);
	return bmp;
}

ok that was if you wanted to have exactly the same bytes as in the memory There are two other ways I can think of. Note that these won't produce the exact copy of the bitmap bits from the memory

Method 2: using a memory stream
This just asks the .Save() method to save the bytes to a memory stream, then you can get the written bytes to the stream:

// Bitmap bytes have to be created via a direct memory copy of the bitmap
private byte[] BmpToBytes_MemStream (Bitmap bmp)
{
	MemoryStream ms = new MemoryStream();
	// Save to memory using the Jpeg format
	bmp.Save (ms, ImageFormat.Jpeg);
	
	// read to end
	byte[] bmpBytes = ms.GetBuffer();
	bmp.Dispose();
	ms.Close();

	return bmpBytes;
}

converting that back to a bitmap is easy.

//Bitmap bytes have to be created using Image.Save()
private Image BytesToImg (byte[] bmpBytes)
{
	MemoryStream ms = new MemoryStream(bmpBytes);
	Image img = Image.FromStream(ms);
	// Do NOT close the stream!
	
	return img;
}

Method 3:using serialization
I can't see any reason why someone would use this method, it's very stupid including it just for reference

// import these
// using System.Runtime.Serialization;
// using System.Runtime.Serialization.Formatters.Binary;
private byte[] BmpToBytes_Serialization (Bitmap bmp)
{
	// stream to save the bitmap to
	MemoryStream ms = new MemoryStream();
	BinaryFormatter bf = new BinaryFormatter();
	bf.Serialize (ms, bmp);

	// read to end
	byte[] bmpBytes = ms.GetBuffer();
	bmp.Dispose();
	ms.Close();

	return bmpBytes;
}
private Bitmap BytesToBmp_Serialized (byte[] bmpBytes)
{
	BinaryFormatter bf = new BinaryFormatter ();
	// copy the bytes to the memory
	MemoryStream ms = new MemoryStream (bmpBytes);
	return (Bitmap)bf.Deserialize(ms);
}

big note: if you use any of these methods, then you have to use both functions presented in the method together. For example you can't get the bitmap bytes using method 1, and then call method2's function BytesToImg()

also, a bit after I wrote that method 1, I realized that there was an API for doing this . It's maybe the best method but there is a turn down: you have to use reflection to retreive the bitmap handle in order to use it with the api function:
Method 4: using reflection and API
I just tell you the APIs to retreive the bitmap handle you can use this

		private IntPtr getImageHandle (Image img)
		{
			FieldInfo fi = typeof(Image).GetField("nativeImage", BindingFlags.NonPublic | BindingFlags.Instance);
			if (fi == null)
				return IntPtr.Zero;

			return (IntPtr)fi.GetValue (img);
		}

the GetBitmapBits API seems to be the easy way out, but apparently it's just there for 16bit compatibility. You should use the GetDIBits instead. If you end up using that method, please post your code




it is your sacred oblication to study these methods and write back ... critisize/suggest better ways

[wossname]

I use Method 1 (with optimisations). All the others are a waste of time

If you use 32bppARGB (or 32bppPARGB like I do) then you can use Marshal.Copy(int[]...) instead which is marginally faster apparently (I've not tested the difference).

I personally dislike using 24bit bitmaps when I'm doing unsafe code, having the stride value different to the width of the image is ugly. Not to mention costly in performance terms. Since 32bpp is more CPU friendly it tends to run better in my experience. Were using all 32 bit color in D# because we need the transparency capability (see the latest screenies in the DoomSharp thread).

We'll be converting all our textures to int arrays soon so we can get a performance boost (if i ever manage to find the time to do this blitter).

[MrPolite]

Originally Posted by wossname

I use Method 1 (with optimisations). All the others are a waste of time

If you use 32bppARGB (or 32bppPARGB like I do) then you can use Marshal.Copy(int[]...) instead which is marginally faster apparently (I've not tested the difference).

I personally dislike using 24bit bitmaps when I'm doing unsafe code, having the stride value different to the width of the image is ugly. Not to mention costly in performance terms. Since 32bpp is more CPU friendly it tends to run better in my experience. Were using all 32 bit color in D# because we need the transparency capability (see the latest screenies in the DoomSharp thread).

We'll be converting all our textures to int arrays soon so we can get a performance boost (if i ever manage to find the time to do this blitter).

I see. I use 32bit in my own app but I thought 24 would be more appropriate for this example

[drepin]

For .NET Framework 2.0 the following line of code (C#) seems to do the trick:

byte[] bytes = (byte[])TypeDescriptor.GetConverter(bmp).ConvertTo(bmp, typeof(byte[]));

Dmitriy

[MrPolite]

interesting! thanks!

[mattes]

I know this post is old, but I'm trying to get an answer to my unexplained problem...
I use your method 1 to manipulate a bitmap.
I create an array with the datas of the original bitmap, make some calculation on it, and save the result in another array, wich i save to a.bmp using the second part of your method.
The problem is if the ratio H/W of the final image is a decimal number (not real) the output image is perfect, if this ratio is real the final array has the same number of cells (x3) as the number of pixel of the final image but produced actually a shifted image, juste like instead of writing 3*W cells per line, it skips one and writes actually 3*W cells +1 (but one is skipped so it writes 3*W cells but 3*W cells shifted by one cell from the 3*W cells it should have written).
Example :

image width = 3 pixels (24Bpp)
array length for one line 9 cells
what it should write : [123456789]1
what it actually writes : 1[234567891]
and do so for each line, one cell is skipped per line