PT Journal
AU Albert Gordo
   Florent Perronnin
   Ernest Valveny
TI Large-scale document image retrieval and classification with runlength histograms and binary embeddings
SO Pattern Recognition
JI PR
PY 2013
BP 1898
EP 1905
VL 46
IS 7
DI 10.1016/j.patcog.2012.12.004
DE visual document descriptor; compression; large-scale; retrieval; classification
AB We present a new document image descriptor based on multi-scale runlengthhistograms. This descriptor does not rely on layout analysis and can becomputed efficiently. We show how this descriptor can achieve state-of-theartresults on two very different public datasets in classification and retrievaltasks. Moreover, we show how we can compress and binarize these descriptorsto make them suitable for large-scale applications. We can achieve state-ofthe-art results in classification using binary descriptors of as few as 16 to 64bits.
ER