image segmentation(to be continue)

canny：

canny's approach is based on three basic objectives:

1.low error rate.All eadge should be found,and there should be no spurious responses.

2.Edge points should be well localized.

3.Single edge point response.

the canny edge detection algorithm consists of the following basic steps:

1.Smooth the input image with a Gaussian filter

2.Compute the amplitude and direction of gradient by using the finite difference of the first order partial derivative

3.Apply nonmaxima suppression to the gradient amplitude image

4.Use double threshholding and connectivity analysis to detect and link edge.