如何采样轮廓序列

Question

I use opencv sobel filter to find the image's contour.

Now I want to sampling the contour's sequence such like {(0,1),(2,2),(3,4)...}.

I need the algorithm can automatic separate each curve.

Is there any algorithm in opencv or I have to do it by myself?

I only think a simple way that scan the image and record the white point by deep first search but I think it performance isn't good.

Answer 1

What's wrong with findContours ?

Given an edge image, the purpose of findContours is to give you a two lists of structures, where one list gives you the pixel coordinates of each contour, and the other list gives a variety of pertinent information that belongs to each detected contour.

In C++, it's as simple as the following. Assuming your image is stored in a cv::Mat structure, you would do something like:

#include "opencv2/imgproc/imgproc.hpp"
using namespace cv;

Mat edge_output;  // Declare edge image
// ...
// ...
// Perform sobel detector on image and store in edge_output
// ...
//

// Declare a vector of Points and a vector to store the hierarchy
vector<vector<Point> > contours;
vector<Vec4i> hierarchy;

// Find contours
findContours(edge_output, contours, hierarchy, CV_RETR_TREE, CV_CHAIN_APPROX_NONE);

hierarchy contains information about the image topology. You can read this page on what the hierarchy means in more detail: http://docs.opencv.org/trunk/doc/py_tutorials/py_imgproc/py_contours/py_contours_hierarchy/py_contours_hierarchy.html

However, what you want to focus on is contours . It would be a vector containing vectors of points. Each element in contours would give you a list of Point structures that contain the (x,y) coordinates of a particular contour. CV_RETR_TREE gives you the full information of each contour, and CV_CHAIN_APPROX_NONE gives you all possible points along the contour.