我们给出一个模板 和一幅图象 。不难发现原图中左边暗，右边亮，中间存在着一条明显的边界。进行模板操作后的结果如下： 。



可以看出，第3、4列比其他列的灰度值高很多，人眼观察时，就能发现一条很明显的亮边，其它区域都很暗，这样就起到了边沿检测的作用。



为什么会这样呢？仔细看看那个模板就明白了，它的意思是将右邻点的灰度值减左邻点的灰度值作为该点的灰度值。在灰度相近的区域内，这么做的结果使得该点的灰度值接近于0；而在边界附近，灰度值有明显的跳变，这么做的结果使得该点的灰度值很大，这样就出现了上面的结果。



这种模板就是一种边沿检测器，它在数学上的涵义是一种基于梯度的滤波器，又称边沿算子，你没有必要知道梯度的确切涵义，只要有这个概念就可以了。梯度是有方向的，和边沿的方向总是正交(垂直)的，例如，对于上面那幅图象的转置图象，边是水平方向的，我们可以用梯度是垂直方向的模板 检测它的边沿。



例如，一个梯度为45度方向模板 ，可以检测出135度方向的边沿。-we give a template and an image. It is not difficult to find the maximum were left dark, right-liang, in the middle there is a clear boundary. After the template for the operation results are as follows :. Can be seen, three, four out other than the gray value is much higher, eye observation, we can obviously found a bright side. Other regions are dark, and this has played a role in the detection of 2500. Why is this the case? A closer look at the template on which to understand it. It means the right to the point o gray minus left point as a gray value of the point of gray values. In a similar gray area, do so as a result of the point of gray values close to 0; And near the border. gray values jump significantly changed, the results do make the point very gray value, and this appeared to