Form document processing is an important topic in office automation area. It solves the problem of analyzing images of form documents automatically. A form processing system takes form document images as input. It will analyze and identify the form structural components and present them back to users in the way of human-perceptible meaning of form documents. In this thesis, I will give an overview of document understanding techniques, especially those that are used in form document processing. I will also propose a form processing algorithm that base on crossing point detection, text block extraction and frame line recovery. An experimental form document processing system will also be presented as the implementation of this thesis.