Visibly accurate model-based binary image compression scheme

In this paper we propose a model-based binary image compression scheme. In this scheme, we merge one-dimensional (1-D) blocks of black pixels of the input binary image with those in consecutive rows into larger blocks using mathematical models that preserve the quality of the image. This process reduces the number of vertices when the image is segmented into rectangles for compression. The top-left and the bottom-right vertices of each generated rectangle are then identified and the coordinates of which are efficiently encoded. The model for merging the blocks was obtained through extracting the data values involving the blocks of various widths and the subjective tolerance of an average viewer. The data values are then plotted on a Cartesian plane and approximated with linear, logarithmic, and polynomial functions. Simulation results show that the images after merging using the proposed model have less number of rectangles without any obvious image distortion and have higher compression ratio those rectangular partitioning methods in the literature.