In this paper, we propose a near minimum sparse pattern coding based scheme for binary image compression. Sparse patterns such as those obtained from prediction, image differencing, and other methods can be coded efficiently using the scheme proposed in this paper. In this research we apply our scheme on coordinate representation of rectangular regions via a number of matrices. Such representations allow for efficiently coding these vertices, and hence compress the image significantly. Simulation results show that the proposed scheme outperformed previously published methods for coordinate data coding by nearly 17%. This scheme has low complexity compared with JBIG2.