Issues When Manipulating Images

Although displaying images is very simple, it still pays to have some understanding of what’s going on behind the scenes.

The most important point to understand about images is that they are always rectangular. That’s not just a convenience, it’s because of the underlying technology. All modern graphics cards have hardware built in that can efficiently copy blocks of pixels from one area of memory to another area of memory, provided that the block of pixels represents a rectangular region. This hardware-accelerated operation can occur virtually as one single operation, and as such is extremely fast. Indeed, it is the key to modern high-performance graphics. This operation is known as a bitmap block transfer (or BitBlt). Graphics .DrawImageUnscaled() internally uses a BitBlt, which is why you can see a huge image, perhaps containing as many as a million pixels, appearing almost instantly. If the computer had to copy the image to the screen pixel by pixel, you’d see the image gradually being drawn over a period of up to several seconds.

BitBlts are very efficient; therefore almost all drawing and manipulation of images is carried out using them. Even some editing of images will be done by manipulating portions of images with BitBlts between DCs that represent areas of memory. In the days of GDI, the Windows 32 API function BitBlt() was arguably the most important and widely used function for image manipulation, though with GDI+ the BitBlt operations are largely hidden by the GDI+ object model.

It’s not possible to BitBlt areas of images that are not rectangular, although similar effects can be easily simulated. One way is to mark a certain color as transparent for the purposes of a BitBlt, so that areas of that color in the source image will not overwrite the existing color of the corresponding pixel in the destination device. It is also possible to specify that in the process of a BitBlt, each pixel of the resultant image will be formed by some logical operation (such as a bitwise AND) on the colors of that pixel in the source image and in the destination device before the BitBlt. Such operations are supported by hardware acceleration and can be used to give a variety of subtle effects. Note that the Graphics object implements another method, DrawImage(). This is similar to DrawImageUnscaled() but comes in a large number of overloads that allow you to specify more complex forms of BitBlt to be used in the drawing process. DrawImage() also allows you to draw (using BitBlt) only a specified part of the image, or to perform certain other operations on it such as scaling it (expanding or reducing it in size) as it is drawn.