I've been using GIST, HOG and SURF descriptors for extracting features from different collections of Chest-X-rays. I could repeatedly see that one descriptor performs better than the other and the performance is not the same across these collections though all are frontal chest X-rays. What attributes to these differences in performance across the collections though they are from the same imaging modality?

More Sivaramakrishnan Rajaraman's questions See All
Similar questions and discussions