AbstractThis paper proposes an encoding scheme based on the concept of vector quantization, an encoding method for storing a large amount of digital line‐drawings. The vector quantization is a well‐known concept as a highly efficient encoding method for speech and image at low transmission rate. In time‐series data such as speech, the mean‐square distortion per sample is employed as the measure for the distortion in quantization, while in the line‐drawing, it is difficult clearly to specify the correspondence to the sample points. This paper proposes a matching by dynamic programming and demonstrates its validity. From the viewpoint that the distortion in encoding depends greatly on the connections among vector patterns, an encoding method permitting the encoding delay is proposed. It is shown that the distortion‐rate performance is improved drastically. The method is applied to a geographical map, and it was verified that a satisfactory figure is reproduced by approximately 1.0 bit per