一种基于QC距离的形状匹配方法

A QC-Distance Based Shape Matching Approach

形状上下文特征描述是图像描述的一种,形状特征向量的提取和相似度量方法至关重要.提出了一种基于QC距离的形状匹配方法,首先计算所有形状轮廓点间径向和角度的相对位置关系,得到形状的特征描述分布,然后利用QC距离来度量特征分布间的相似度,最终实现形状匹配.在手写数字公测数据集MNIST上的实验结果表明,引入QC距离度量有效地提高了匹配的准确率.

Shape context is a classical approach of image description. Constructing the shape feature vectors and the distance measurement method between two vectors are both important. This paper has presented a QC-distance based shape matching approach. Firstly, the relative positional relationship has been calculated between the radial and angular shape contour points, which give shape feature distribution. Secondly, QC is used to measure the distance between two shape feature distributions. Finally, the shape matching has been found through optimization method. Experimental results on handwritten Handwritten Digital Library MINST show that employing the QC-distance for shape matching can offer better performance.