OpenCV  4.2.0
Open Source Computer Vision
Canny Edge Detector

Prev Tutorial: Laplace Operator
Next Tutorial: Hough Line Transform

Goal

In this tutorial you will learn how to:

  • Use the OpenCV function cv::Canny to implement the Canny Edge Detector.

Theory

The Canny Edge detector [Canny86] was developed by John F. Canny in 1986. Also known to many as the optimal detector, the Canny algorithm aims to satisfy three main criteria:

  • Low error rate: Meaning a good detection of only existent edges.
  • Good localization: The distance between edge pixels detected and real edge pixels have to be minimized.
  • Minimal response: Only one detector response per edge.

Steps

  1. Filter out any noise. The Gaussian filter is used for this purpose. An example of a Gaussian kernel of \(size = 5\) that might be used is shown below:

    \[K = \dfrac{1}{159}\begin{bmatrix} 2 & 4 & 5 & 4 & 2 \\ 4 & 9 & 12 & 9 & 4 \\ 5 & 12 & 15 & 12 & 5 \\ 4 & 9 & 12 & 9 & 4 \\ 2 & 4 & 5 & 4 & 2 \end{bmatrix}\]

  2. Find the intensity gradient of the image. For this, we follow a procedure analogous to Sobel:
    1. Apply a pair of convolution masks (in \(x\) and \(y\) directions:

      \[G_{x} = \begin{bmatrix} -1 & 0 & +1 \\ -2 & 0 & +2 \\ -1 & 0 & +1 \end{bmatrix}\]

      \[G_{y} = \begin{bmatrix} -1 & -2 & -1 \\ 0 & 0 & 0 \\ +1 & +2 & +1 \end{bmatrix}\]

    2. Find the gradient strength and direction with:

      \[\begin{array}{l} G = \sqrt{ G_{x}^{2} + G_{y}^{2} } \\ \theta = \arctan(\dfrac{ G_{y} }{ G_{x} }) \end{array}\]

      The direction is rounded to one of four possible angles (namely 0, 45, 90 or 135)
  3. Non-maximum suppression is applied. This removes pixels that are not considered to be part of an edge. Hence, only thin lines (candidate edges) will remain.
  4. Hysteresis: The final step. Canny does use two thresholds (upper and lower):

    1. If a pixel gradient is higher than the upper threshold, the pixel is accepted as an edge
    2. If a pixel gradient value is below the lower threshold, then it is rejected.
    3. If the pixel gradient is between the two thresholds, then it will be accepted only if it is connected to a pixel that is above the upper threshold.

    Canny recommended a upper:lower ratio between 2:1 and 3:1.

  5. For more details, you can always consult your favorite Computer Vision book.

Code

  • What does this program do?
    • Asks the user to enter a numerical value to set the lower threshold for our Canny Edge Detector (by means of a Trackbar).
    • Applies the Canny Detector and generates a mask (bright lines representing the edges on a black background).
    • Applies the mask obtained on the original image and display it in a window.
  • Explanation (C++ code)

    1. Create some needed variables:
      Mat src, src_gray;
      Mat dst, detected_edges;
      int lowThreshold = 0;
      const int max_lowThreshold = 100;
      const int ratio = 3;
      const int kernel_size = 3;
      const char* window_name = "Edge Map";
      Note the following:
      1. We establish a ratio of lower:upper threshold of 3:1 (with the variable ratio).
      2. We set the kernel size of \(3\) (for the Sobel operations to be performed internally by the Canny function).
      3. We set a maximum value for the lower Threshold of \(100\).
    2. Loads the source image:
      CommandLineParser parser( argc, argv, "{@input | fruits.jpg | input image}" );
      src = imread( samples::findFile( parser.get<String>( "@input" ) ), IMREAD_COLOR ); // Load an image
      if( src.empty() )
      {
      std::cout << "Could not open or find the image!\n" << std::endl;
      std::cout << "Usage: " << argv[0] << " <Input image>" << std::endl;
      return -1;
      }
    3. Create a matrix of the same type and size of src (to be dst):
      dst.create( src.size(), src.type() );
    4. Convert the image to grayscale (using the function cv::cvtColor ):
      cvtColor( src, src_gray, COLOR_BGR2GRAY );
    5. Create a window to display the results:
      namedWindow( window_name, WINDOW_AUTOSIZE );
    6. Create a Trackbar for the user to enter the lower threshold for our Canny detector:
      createTrackbar( "Min Threshold:", window_name, &lowThreshold, max_lowThreshold, CannyThreshold );
      Observe the following:
      1. The variable to be controlled by the Trackbar is lowThreshold with a limit of max_lowThreshold (which we set to 100 previously)
      2. Each time the Trackbar registers an action, the callback function CannyThreshold will be invoked.
    7. Let's check the CannyThreshold function, step by step:
      1. First, we blur the image with a filter of kernel size 3:
        blur( src_gray, detected_edges, Size(3,3) );
      2. Second, we apply the OpenCV function cv::Canny :
        Canny( detected_edges, detected_edges, lowThreshold, lowThreshold*ratio, kernel_size );
        where the arguments are:
        • detected_edges: Source image, grayscale
        • detected_edges: Output of the detector (can be the same as the input)
        • lowThreshold: The value entered by the user moving the Trackbar
        • highThreshold: Set in the program as three times the lower threshold (following Canny's recommendation)
        • kernel_size: We defined it to be 3 (the size of the Sobel kernel to be used internally)
    8. We fill a dst image with zeros (meaning the image is completely black).
      dst = Scalar::all(0);
    9. Finally, we will use the function cv::Mat::copyTo to map only the areas of the image that are identified as edges (on a black background). cv::Mat::copyTo copy the src image onto dst. However, it will only copy the pixels in the locations where they have non-zero values. Since the output of the Canny detector is the edge contours on a black background, the resulting dst will be black in all the area but the detected edges.
      src.copyTo( dst, detected_edges);
    10. We display our result:
      imshow( window_name, dst );

    Result

    • After compiling the code above, we can run it giving as argument the path to an image. For example, using as an input the following image:

    • Moving the slider, trying different threshold, we obtain the following result:

    • Notice how the image is superposed to the black background on the edge regions.
    cv::String
    std::string String
    Definition: cvstd.hpp:150
    cv::Scalar_< double >::all
    static Scalar_< double > all(double v0)
    returns a scalar with all elements set to v0
    cv::IMREAD_COLOR
    @ IMREAD_COLOR
    If set, always convert image to the 3 channel BGR color image.
    Definition: imgcodecs.hpp:67
    cv::cvtColor
    void cvtColor(InputArray src, OutputArray dst, int code, int dstCn=0)
    Converts an image from one color space to another.
    cv::samples::findFile
    cv::String findFile(const cv::String &relative_path, bool required=true, bool silentMode=false)
    Try to find requested data file.
    cv::waitKey
    int waitKey(int delay=0)
    Waits for a pressed key.
    cv::Canny
    void Canny(InputArray image, OutputArray edges, double threshold1, double threshold2, int apertureSize=3, bool L2gradient=false)
    Finds edges in an image using the Canny algorithm .
    highgui.hpp
    cv::namedWindow
    void namedWindow(const String &winname, int flags=WINDOW_AUTOSIZE)
    Creates a window.
    cv::Size
    Size2i Size
    Definition: types.hpp:347
    cv::Canny
    void Canny(InputArray dx, InputArray dy, OutputArray edges, double threshold1, double threshold2, bool L2gradient=false)
    cv::imread
    Mat imread(const String &filename, int flags=IMREAD_COLOR)
    Loads an image from a file.
    cv::Mat::empty
    bool empty() const
    Returns true if the array has no elements.
    cv::Mat::size
    MatSize size
    Definition: mat.hpp:2108
    cv::dnn::print
    static void print(const MatShape &shape, const String &name="")
    Definition: shape_utils.hpp:188
    cv::imshow
    void imshow(const String &winname, InputArray mat)
    Displays an image in the specified window.
    cv::Scalar
    Scalar_< double > Scalar
    Definition: types.hpp:669
    cv::blur
    void blur(InputArray src, OutputArray dst, Size ksize, Point anchor=Point(-1,-1), int borderType=BORDER_DEFAULT)
    Blurs an image using the normalized box filter.
    cv::Mat
    n-dimensional dense array class
    Definition: mat.hpp:792
    cv::imshow
    void imshow(const String &winname, const ogl::Texture2D &tex)
    Displays OpenGL 2D texture in the specified window.
    cv::Mat::type
    int type() const
    Returns the type of a matrix element.
    cv::Mat::copyTo
    void copyTo(OutputArray m) const
    Copies the matrix to another one.
    cv::CommandLineParser
    Designed for command line parsing.
    Definition: utility.hpp:797
    cv::COLOR_BGR2GRAY
    @ COLOR_BGR2GRAY
    convert between RGB/BGR and grayscale, color conversions
    Definition: imgproc.hpp:542
    cv::createTrackbar
    int createTrackbar(const String &trackbarname, const String &winname, int *value, int count, TrackbarCallback onChange=0, void *userdata=0)
    Creates a trackbar and attaches it to the specified window.
    cv
    "black box" representation of the file storage associated with a file on disk.
    Definition: affine.hpp:52
    imgproc.hpp
    cv::WINDOW_AUTOSIZE
    @ WINDOW_AUTOSIZE
    the user cannot resize the window, the size is constrainted by the image displayed.
    Definition: highgui.hpp:184
    cv::Mat::create
    void create(int rows, int cols, int type)
    Allocates new array data if needed.