Based on an analysis of the existing text watermarking techniques, a high-capacity text watermarking scheme resistive to print-and-scan processes is proposed. The binary document image is first segmented into individual characters. A multi-bit watermark is then embedded by flipping the boundary pixels of each character based on a print-scan invariant quantity and complexity of the character. A quantitative model is established to achieve the character pixel quantization, which can minimize visual distortion. Experimental results show that the proposed scheme can increase embedding capacity. It can effectively resist print-scan attacks so as to watermark hardcopy documents.