Reading GUIs is a lot easier than most things tesseract was designed for. You may still need a little preprocessing; I suggest OpenCV.

Basically enlarge it, filter to remove noise (sharpening and blob detection and perhaps another), threshold to get BW image.

Most unneeded for GUI. OpenCV helps greatly in the situations Schutz describes. Modification of the character data is better but more time intensive.