You can certainly produce graphics that include polygons and text and then output the results using GD. You can also load an existing image into GD and copy a sub-rectangle of it to a new image. Ie crop. I don't think it has any facilities for scaling.

The salient question here is what format are those "existing graphics" in? If you mean images in the form of jpg or png or similar, then GD can load and modify (by over writing) and output those. If the graphics are in some vector graphic format you would need to render them to an image format first.