MRF随机场在文本图像清晰化算法中的应用

Enhancement of Document Text Images Using MRF Approach

对于扫描或相机拍摄的低分辨率文本图片,提出一种基于马尔科夫随机场的文本图像清晰化算法。该算法首先引入了马尔科夫随机场模型,然后考虑平衡数据项与平滑项之间的关系从而确定出能量函数的表达式。利用改进过的置信度传播方法来解决马尔科夫随机场中的最大后验估计问题,以达到最小化能量函数的目的。实验结果表明,本算法能够较有效地提高低分辨率图像的清晰度,一定程度上提高了图片的分辨率,尤其对文字的边缘部分有很好的锐化效果。

Document text images captured by book scanner or camera usually require resolution enhancement for other usage,present a text image enhancement algorithm based on Markov random field.Firstly the algorithm introduces the MRF model,and then defines the energy function considering the relation between data constraint and smoothness constraint.The improved belief propagation was implemented to resolve the MAP（Maximum A Posterior） of MRF model,in order to minimize the energy function.The experiment result shows that the algorithm can efficiently enhance the low resolution text image and improve the image resolution to some extent.And the edge of the text has a good sharpening effect.