I am using sift feature extractor for image stitching . using vl_sift in matlab. but the extracted features are not distributed evenly in pictures . in some point I have several of them while other points I don't have any . how can I avoid this ?
Very simply you can just run the algorithm, lower any outlier thresholds you're using to guarantee you have more points everywhere, then manually delete points from regions where there are many.
{however this is really not recommended at all}