一种全局和局部相结合的二值化方法研究

为了得到清晰准确的二值化图像，且避免全局阈值化对图像细微信息的二值化不清晰和局部阈值二值化中往往存在伪影现象，提出了一种借助Candy算子检测边缘特征，将改进的二维Otsu最佳全局阈值与局部Bernsen阈值法在边缘信息的基础上相结合的二值化方法。实验表明这种方法能很好地处理带有噪声和低对比度的图像，通过改进的二维Otsu法降低了传统二维Otsu法的时间复杂度，并且可以减少局部阈值法的块效应，避免了大量虚假信息，又能保留图像的细节信息。

An Improved Binarization Method Combined Global Thresholding with Local Thresholding

In order to obtain clear accurate binarization image, avoiding the binarizafion not clear of image subtle information by global thresholding and artifact phenomenon existed in local thresholding, a binarization method combined improved 2d-Otsu best global thresh- olding with local Berasen thresholding based on edge information is proposed ,detecting edge feature with the aid of Candy operator. The experiments indicate that the method can handle with image with noise and low contrast well ,reduce the time complexity of traditional 2d Otsu by improved 2d-Otsu and block effect of local thresholding,avoid the much false information and keep the detail of the image in formation.