This might be due to a large page size and/or small zoom level. See http://forums.fofou.org/sumatrapdf/topic?id=99347 .
An image stream has a corrupt never-ending flate encoded stream. When we try to load it all, we naturally run out of memory... Another bug exists to be smarter about loading image data on an as needed basis. When that bug is fixed, we won't crash on this one.