This paper deals with two classes of preprocessing techniques for documents which are essentially black and white. These are bilevel quantization and preprocessing techniques for noisy input documents. The effect of such techniques on the compression ratio and the subjective quality of the output copy are studied. It is shown that for noisy documents, certain preprocessing techniques considerably improve the compression ratio. The improvement is substantially more for two-dimensional coding algorithm such as relative address coding (RAC). Simulation results of the effect of preprocessing techniques for noise reduction for both multilevel and bilevel signals are described The results of a subjective experiment to determine the effect of certain preprocessing techniques on the quality of clean and noisy documents are reported.