基于最小均方预测的图像单像素精度边缘检测算法

基于预测的思想，引入数字信号处理中常用的最小均方（LMS）横向滤波器对图像进行边缘检测。由于图像的缓变区域可等效为一个平稳随机过程，使得LMS算法能够适用。在通过LMS预测检测出图像边缘的基础上，本文还对该算法在边缘处的行为进行了附加条件的判断，使得LMS算法在经过边缘的地方重新收敛后才进行边缘检测，这样可以获得单像素精度边缘。

Image single-pixel-precision edge detection algorithm based on LMS prediction

To gain better effect of edge detection, an algorithm based on the idea of prediction was proposed with the help of LMS adaptive transversal filter theory which was commonly used in digital signal processing. Because the low frequency area of one image can be modeled as a stationary stochastic process, the LMS algorithm could be adopted. Besides edge detection, we also added an auxiliary criterion which could assure that the LMS algorithm have attained convergence after the oscillation caused by the edge encounter, This criterion can lead to single-pixel-precision results which was one important object of edge detection.