Selecting keypoints in images using descriptor scores

A method comprising:

• detecting, via a processor, first keypoints in a plurality of images;
• calculating, via the processor, a keypoint score for each of the first keypoints based on a weighted combination of a descriptor score and a detector score, the descriptor score indicating descriptor invariance, the detector score indicating magnitudes of gradients around the detected first keypoints;
• selecting, via the processor, second keypoints from the first keypoints based on the calculated keypoint scores;
• calculating, via the processor, descriptors for each of selected second keypoints;
• matching, via the processor, descriptors between in the plurality of the images;
• and
• tracking, via the processor, a feature in the plurality of the images based on the matched descriptors.

At least one non-transitory computer readable medium having instructions stored therein that, in response to being executed on a computing device, cause the computing device to:

• detect first keypoints in a plurality of images;
• calculate a keypoint score for each of the first keypoints based on a weighted combination of a descriptor score and a detector score, the descriptor score indicating descriptor invariance, the detector score indicating magnitudes of gradients around the detected first keypoints;
• select second keypoints from the first keypoints based on the calculated keypoint scores;
• calculate descriptors for each of the second keypoints;
• match corresponding descriptors between the plurality of the images;
• and
• track a feature in the plurality of the images based on the matched corresponding descriptors.