Revision history [back]

The coordinates of a keypoint are saved with floating values. So the coordinates do not exactly match with a certain pixel but lies within one pixel. That is what I unterstood so far and what I've been told. I have searched the internet to find out how OpenCv determines the color of a keypoint. Sadfully, I haven't found any answers.

Is OpenCv taking directly the color of the pixel or does OpenCv consider the neighbors, too?

There are some missunderstandings, so I changed the question:

The coordinates of a keypoint are saved with floating values.

KeyPoint(float x, float y, float _size, float _angle=-1, float _response=0, int _octave=0, int _class_id=-1)

So the coordinates do not exactly match with a certain pixel but lies is within one pixel. That is what That's how I unterstood so far and what I've been told. unterstand it. Maybe I missunderstood.

I detect features to get color information. When I have searched keypointBob he is somewhere in the internet to find out how OpenCv determines image img. I can get the color of a keypoint. Sadfully, I haven't found any answers.value by calling: img.at<vec3b>( keypointBob.pt)[]

Is OpenCv taking directly the color of the pixel (where the keypoint is within) or does OpenCv consider the neighbors, too? neighbors?

I hope it's more clear now.