Pattern Matching is the most conventional method of binary text image compression that has been only used in the 2-D domain of textual image signals. In this paper a pattern matching technique is proposed in the 1-D domain of chain code description signal of printed binary textual Farsi-Arabic images. In printed Farsi-Arabic scripts, contrary to latin scripts, letters usually attach to each other and produce many different patterns. Hence some patterns are fully or partially subsets of others. Detecting such situations and exploiting them to reduce the number of library prototypes has a great effect on the compression efficiency. The Proposed method, contrary to the existing compression methods, has used this property for increasing the compression ratio. For the template matching part of the proposed method, we may use either the cross correlation or a proposed similarity measure which has lower computation time and better results. Experimental results show that the compression performance of the proposed method is as high as 4.5 times that of the conventional one.