A Novel Multiscale Edge Detection Approach Based on Nonsubsampled Contourlet Transform and Edge Tracking

Algorithm 1

Input: the set of edge images and the set of gradient orientation maps .

Output: the final edge tracking map .

Begin

Set , where is currently tracking scale variable and is decomposition level of the NSCT.

While  

(1) Remove isolated pixels for edge image .

(2) Get gradient direction for each pixel , where .

(3) Form the set of edge pixels for each pixel corresponds to ,

    where are these pixels satisfying: the maximum displacement in direction

     and . If , then and .

(4) Generate a strong edge map and a weak edge map for and ,

    respectively. Then perform edge connector for and , as is described in Section 3.3, the

    result is assigned to .

(5) Perform the operation and go back to (1).

End

.

End