KeyPoint descriptor OpenCV

Question

I am trying to understand how to get the descriptor for a given KeyPoint in OpenCV. So far my code looks like follows:

#include <iostream>
#include "opencv2/opencv.hpp"

typedef cv::Mat Image;

int main(int argc, const char * argv[])
{

    Image imgA = cv::imread("images/buddhamulticam_total100.png",
                              CV_LOAD_IMAGE_GRAYSCALE);
    Image imgB = cv::imread("images/buddhamulticam_total101.png", 
                            CV_LOAD_IMAGE_GRAYSCALE);

    cv::Ptr<cv::FeatureDetector> detector = 
                    cv::FeatureDetector::create("ORB");
    cv::Ptr<cv::DescriptorExtractor> descriptor = 
                    cv::DescriptorExtractor::create("ORB");

    std::vector<cv::KeyPoint> keyPointsA, keyPointsB;

    keyPointsA.push_back(cv::KeyPoint(0,0,5));
    keyPointsB.push_back(cv::KeyPoint(10,10,5));

    cv::Mat descriptorA, descriptorB;
    descriptor->compute(imgA, keyPointsA, descriptorA);
    descriptor->compute(imgB, keyPointsB, descriptorB);

    std::cout << "DescriptorA (" << descriptorA.rows << "," <<
                descriptorA.cols << ")" << std::endl;
    std::cout << "DescriptorB (" << descriptorB.rows << "," 
              << descriptorB.cols << ")" << std::endl;
    return 0;
}

The problem is that I am getting no data in the descriptor. What am I missing? Could you explain in more detail what are the params passed to the KeyPoint object? I am new to computer vision + OpenCV, so probably a better explanation (than OpenCV's documentation) could help.

Let me understand, the points are (0,0) and (10,10) ? – GilLevi Oct 19 '13 at 22:56 — GilLevi, Oct 19 '13 at 22:56
Yes, that's what opencv's doc says. – BRabbit27 Oct 20 '13 at 05:26 — BRabbit27, Oct 20 '13 at 05:26

GilLevi · Accepted Answer · 2013-10-20T16:06:57.240

2

You're trying to computer ORB on the points (0,0) and (10,10), but they are too close to the image border, so ORB can't compute descriptors in those locations. ORB (as well as the other binary descriptors) filters them out.

EDIT: since you asked about usage, I'm editing the answer. You should pass the whole image. I use it as:

Ptr<FeatureDetector> detector = FeatureDetector::create(detector_name);
Ptr<DescriptorExtractor> descriptor = DescriptorExtractor::create(descriptor_name);

detector->detect(imgK, kp);
descriptor->compute(imgK, kp, desc);

edited Oct 20 '13 at 16:06

answered Oct 20 '13 at 10:02

GilLevi

2,117
5
22
38

Ok I get it, but then the way of creating the `KeyPoint` and pass it to the `DescriptorExtractor` is ok? If I take a small patch of 5x5 in the image, when computing the descriptor should I pass the whole image or only that patch? – BRabbit27 Oct 20 '13 at 11:33

KeyPoint descriptor OpenCV

1 Answers1

Linked