基于改进的Sobel算子边缘检测算法

为了解决传统的Sobel算子算法存在的斜向方向不敏感问题，提出了一种改进的Sobel算法。该算法在Sobel算子的基础上，增加了45°和135° 2个方向模板，并且以斜向边缘为主重新分配了算子模板的权重。算法先将图像转换为灰度图像，然后用改进的Sobel算子得到梯度图像，再采用局部梯度阈值对梯度图像进行细化处理，最后结合斜向边缘方向域值进行二值化，得到边缘图像。仿真实验表明，与传统Sobel算法相比，该算法提高了检测精度，使得边缘细节更丰富、更连续。

Edge Detection Algorithm Based on Improved Sobel Operator

An improved Sobel algorithm is proposed to overcome the disadvantage, which the traditional Sobel operator algorithm is insensitive for the diagonal directions. The algorithm increases two templates in both directions of 45° and 135° ,and reallocates operator template weight according to diagonal edge. The algorithm fast converts the image to grayscale image, and then obtains the gradient image with the improved Sobel operator,then refines the gradient image with local gradient threshold values, finally the algorithm binaries the image according to the direction toward combining selected threshold. Simulation result shows that the algorithm has a higher detection accuracy, achieves more rich and more continuous edge than traditional Sobel algorithm.