双阈值法在文字轮廓提取中的应用

在图像的边缘检测中，利用模角分离的小波变换，结合尺度独立的算法区分了阶梯型边界和跳跃型边界，从而提取了阶梯型边界．上述算法中需选择峰值阈值，将小波变换系数较小的点滤掉，但是一幅图像中边缘的奇异性并不均匀，对变换后的整幅图像取固定阈值时，那么微弱边缘将会随着因灰度不均匀、噪声等一并被滤除．针对这一问题，在上述算法中用双阈值法代替固定阈值，在文字图像轮廓提取中取得了较好的效果．

Application of Double Threshold Value Method in the Abstract of Character Figure

In the image edge detection, the bounds of step-structure and dirac-structure are distinguished by using modular - angle-separated wavelet transform, or combining with scale-independent algorithm, thus obtaining the bounds of the step-structure. In order to filtrate the smaller point of the wavelet coefficient needs to chose threshold value of peak value used former algorithm, However, the singularity of the edge in a image is not uni-form, the faintness edge will be filtered with the non-uniform of gray-scale and noise, etc.. When taking the same threshold from the transformed whole image. To solve this problem, using double threshold value takes the place of character threshold value in the former algorithm. Good effect is achieved in abstract of character figure using the algorithm in this paper.