投影分解方法应用于Canny算法的阈值选取

应用投影分解自适应选择Canny算法的阈值。经过高斯平滑的细胞图像，其梯度直方图呈现多峰特性，而包络呈现双峰特性，为了提取这两个峰，我们应用投影分解将梯度直方图分解为一组高斯分布函数，从而寻找梯度直方图包络中的谷点作为Canny算法的高门限。实验结果表明，新算法较好地解决了传统算法带来的噪声过大或者有用边缘丢失的问题，边缘检测效果得到了改善。

Adaptive Threshold for Canny Operator Based on Projective Decomposition

This paper describes a algorithm of gettng adaptive threshold for canny operator Based on projective Decomposition.After gaussian smoothing,the histogram of the image of gradient appears a multi-peak feature,but it's contour has only two peaks.In order to get these two peaks,we decompose the histogram into a set of gaussian density functions based on projective decomposition.Then the high threshold of canny operator can be calculate with the parameters of these gaussian density function.Comparing with other algorithm.this one has the advantages of better effect in edge detection because of its good performance in suppressing noise and maintaining edges.