Compression of scanned text document images is important in modern document management, communications and retrieval systems. However, most existing compression techniques have been studied extensively only for documents in English or similar alphabet-based languages. In this paper, we purpose a content-lossless scheme for compression of Chinese text documents. This method utilizes the radical characteristics, unique to Chinese characters, to minimize the size of compressed documents. Our method consists of two main parts. The first part is the development of a radical pattern library. The second part is to utilize the radical pattern library to match character patterns in a document. The technique has been tested with many Chinese text document images with good results.