针对传统Sobel边缘检测算法对噪声敏感的缺点进行了分析，提出了自适应权值的改进算法，该权值的大小取决于Sobel算子模板内像素灰度值之和。通过不同形式的噪声图像处理对比，结果表明，该算法不仅计算简单，而且具有较强的抗噪能力，对图像边缘有较高的检验精度，极大地改进了图像边缘识别的效果。

In view of the traditional Sobel algorithm for edge detection is sensitive to noise, an improved Sobel algorithm of adaptive weights is proposed. The weights depend on the sum of pixel gray value within the template of the Sobel operator. In comparison of processing different noise images, the results show that the improved Sobel algorithm is not only in high calculation efficiency, but also with a strong anti-noise ability. It is proved that the improved Sobel algorithm has higher precision and it is much more effective in edge recognition.