AbstractThe development of technology concerning semiconductors and optical fibers now makes image transmission cheaper and more practical. This paper describes an error correcting coding method for binary images of line drawings. First the bilevel images of line drawings are represented by legal patterns, which are defined with a 3 × 3‐pixel unit mesh (window) and employ line constraints. The resulting image is a picture represented by a legal symbol connection (LSC‐picture). This picture includes certain connecting relations among the neighboring legal patterns, which are considered to represent explicitly the redundancy of the original image. Next we encode the picture using a legal pattern as a coding unit. The information source model used for encoding is a simple Markov model. We use line‐to‐line dependencies for the purpose of error detection and correction. When a transmission error is detected by checking the relations, the decoding process generates a tree called an “error correcting tree.” The problem of correcting the transmission error is considered as one of how to search the error correcting tree. The simulation results of this method show that almost all the transmission errors can be corrected if we incorporate a little redundancy of the code words themselves, as in an ordinary method. At the same time, the compression ratio is as high as that of the MR coding method. For image transmission in the future we believe that a description should be considered which can make explicit the redundancy of the image itself as much as possible and that the redundancy should be utilized not only for compressing the amount of data but also for correcting transmi