The character projection (CP) is utilized for maskless lithography and is a potential for the future photomask manufacture because the CP lithography can project ICs much faster than point beam projection or variable-shaped beam (VSB) projection. In this paper, we present CP mask optimization for multi-column-cell (MCC) systems, in which column-cells can project patterns in parallel with the CP and the VSB, so that their throughput is maximized. This paper presents an MINLP (mixed integer nonlinear programming) model as well as an MIP (mixed integer programming) model for optimizing a CP mask set of an MCC projection system so that projection time is minimized. The experimental results show that our optimization has achieved 71.3% less projection time for a two-column-cell system than that for a single-column-cell (SCC) system. For the two-column-cell system, it has also achieved 42.6% less projection time than a naive CP mask development approach. The experimental results denote that our optimization achieves projection time reduction more than parallelizing two column-cells by virtually increasing logic cells which are placed on CP masks and decreasing VSB projection.