采用信号匹配的支票图像二值化

Bank Check Binarization with Signal Matching

提出了一种基于信号匹配的低信噪比图像的信号提取方法,解决类似支票日期域这种既有复杂背景,又有印章噪声干扰的图像二值化问题.这种二值化方法完全不同于传统的方法,它从信号处理的角度出发,利用了部分先验知识和理想状态下的投影轮廓信号,再通过用不同阈值分割的投影信号与之匹配,匹配度最大时的阈值即为图像分割的最佳阈值.本方法比其他的传统的二值化方法更具有自适应性和鲁棒性,通过大量的实验数据,以及与常用的其他算法进行的比较得到了验证.

A method based on signal matching to binarize low signal-noise rate image was proposed. The information from image interference of both complex background and seal noise could be extracted. Making full use of the prior knowledge, the image projection function under ideal condition would be obtained. By using this signal to match the projection functions binarized by different thresholds, the optimum threshold was in fact the threshold which projection function match best. This method is more self-adaptive and robust than other traditional methods. Through experiments, when compared with other algorithms, satisfactory results were got.