The purpose of this paper is to quantitatively evaluate the improvement that can be obtained, in the coding of graphics, by using more sophisticated encoding schemes than run-length. These schemes are now well within the reach of today's technology. In the broad class of encoders by coordinates, three isoentropic encoding schemes are considered: 1) encoding of isolated points, 2) encoding of unidimensional patterns, 3) encoding of two-dimensional patterns. Although the three systems are all conceptually straightforward, they require an increasing complexity of encoding-decoding operations. This complexity is balanced, however, by a larger compression factor for the more complex schemes. The results obtained with these encoding schemes are finally compared to those of a run-length coding scheme for the case of a first-order Markov model of the source.