extract pixels from System.Drawing.Bitmap

Question

I am using some .NET assembly from MATLAB which produces a System.Drawing.Bitmap object. I would like to get a MATLAB matrix with the pixels out of it. How do you do that?

I don't want to save the image to disk and then use imread.

Answer 1

Based on Jeroen's answer, here is the MATLAB code to do the conversion:

% make sure the .NET assembly is loaded
NET.addAssembly('System.Drawing');

% read image from file as Bitmap
bmp = System.Drawing.Bitmap(which('football.jpg'));
w = bmp.Width;
h = bmp.Height;

% lock bitmap into memory for reading
bmpData = bmp.LockBits(System.Drawing.Rectangle(0, 0, w, h), ...
    System.Drawing.Imaging.ImageLockMode.ReadOnly, bmp.PixelFormat);

% get pointer to pixels, and copy RGB values into an array of bytes
num = abs(bmpData.Stride) * h;
bytes = NET.createArray('System.Byte', num);
System.Runtime.InteropServices.Marshal.Copy(bmpData.Scan0, bytes, 0, num);

% unlock bitmap
bmp.UnlockBits(bmpData);

% cleanup
clear bmp bmpData num

% convert to MATLAB image
bytes = uint8(bytes);
img = permute(flipdim(reshape(reshape(bytes,3,w*h)',[w,h,3]),3),[2 1 3]);

% show result
imshow(img)

The last statement can be hard to understand. It is in fact equivalent to the following:

% bitmap RGB values are interleaved: b1,g1,r1,b2,g2,r2,...
% and stored in a row-major order
b = reshape(bytes(1:3:end), [w,h])';
g = reshape(bytes(2:3:end), [w,h])';
r = reshape(bytes(3:3:end), [w,h])';
img = cat(3, r,g,b);

The result:

Answer 2

If you want to modify individual pixels, you can call the Bitmap.SetPixel(..) but this is slow ofcourse.

With the BitmapData you can get the bitmap as array of pixels.

System.Drawing.Imaging.BitmapData bmpData =
            bmp.LockBits(rect, System.Drawing.Imaging.ImageLockMode.ReadWrite,
            bmp.PixelFormat);

IntPtr ptr = bmpData.Scan0;

// code

// Unlock the bits.
bmp.UnlockBits(bmpData);

see: http://msdn.microsoft.com/en-us/library/system.drawing.imaging.bitmapdata.aspx

In the example they use Marshal.Copy, but this is if you want to avoid unsafe.

With unsafe code you can manipulate the pixel data directly.