public class ImageSurface
extends Surface

A Surface which is an image in memory and can be written to disk. If drawing to an image you intend to write to a PNG, you would end up doing something like:

 surface = new ImageSurface(Format.ARGB32, 100, 100);
 cr = new Context(surface);
 
 // do drawing
 
 surface.writeToPNG(filename);
 

While ImageSurfaces are good for writing images out to disk, they are not optimized per se to be efficient as a back end, nor are they accelerated by your graphics card. So they are not an appropriate intermediate in drawing operations; don't be calling setSource() on one of these.

More importantly, ImageSurface is not an image loader! Remember that Surfaces are what Cairo draws to. If what you are doing is building up images for display to the screen, then load your images into XlibSurfaces and use those as sources.

Deep in the guts, Cairo's ImageSurface is like GDK's Pixbuf, a format that C programmers can directly address directly in memory via pointers. That's useful for very low level programming, but not needed for application development. If you're drawing, use Cairo's higher level drawing primitives; if you need to introspect an image, then load it with Pixbuf and use Pixbuf's getPixels() to peek around in its data.