Why some FeatureDetector finds features only in the middle of the image?
I try to make a classifier, and I tested different FeatureDetectors and I have seen some of the detectors find features only in the middle of the image. Like for example STAR:
Is there a possibility to change that, so it finds the feature all over the image, not only in the center region?
EDIT:
The green rectangle is the bounding rect of the centers of the KeyPoints found. Why I do not have KeyPoints outside it, like on the shadow of the car (right), or on the white car, on the top, or between the rocks on the floor. With other images I get the same bounding rect, no displacement. Is it normal that STAR is not searching on the border of the image?
I have used cv::FeatureDetector::create("STAR")
Hmm your image shows features everywhere ... As to finding more features in the center, this is normal due to the number plate having strong edges. The reason why there are no features in the center are quite logical, the black car has no strong edges or pixel intensities changing this they cannot be used as as interesting points.
Ok my istake. I thought that the border was an initial detection of a cascade car classifier. Again I think the same conclusion can be stated. Probably the features at the borders cannot be calculated due to the fact how the feature is calculated. I do not have an exact solution for this, should investigate STAR code further.
A search at the documentation of stardescriptor points me to he function below which adjusts a image boundary parameter. I think that universal featuredetector interface used by both will probably use that parameter to define a border to ignore. Ill keep you posted.
this topic seems to talk about the border functionality also.