基于剪枝信任度传播的手写体目标形状匹配方法

手写体识别中,目标形状的匹配是较为重要的工作.为了提高手写体目标形状的匹配速度,提出一种新的匹配方法.由于手写体目标形状的几何先验知识已知,并可以采用少量的参数进行表示,新方法采用参数化可变形模板匹配目标形状,确定其后验概率模型,并定义剪枝信任度空间,依据信任度传播算法的特性,首次将剪枝信任度传播算法应用于求解可变形模板与目标形状之间的最佳匹配.实验结果显示,在灰度图像中,对手写体目标形状的轮廓检测与定位速度显著提高.提出将剪枝信任度传播方法应用于手写体目标形状的匹配工作,能够使得目标形状填补空白,应用于相关性较为稀疏的图模型中.

The Novel Method of Matching Handwriting's Shapes based on Pruned Belief Propagation

In the study of handwriting's recognition,the matching of shapes is very important.To improve the matching speed,a novel method is presented which can matches handwriting's shapes faster.Owing to the foregone of shapes' geometrical prior probability and a small quantity of parameters,the method matches shapes using parameterized deformable template,and it can denote the posterior probability.Based on the characters of belief propagation,we defined a pruned belief state and it is for the first time that we apply the pruned belief propagation algorithm to find the best match between the template and shape.The experimental results show that the novel method boosts the matching speed in the grayscale image.