Tree-structured Temporal Information for Fast Histogram Computation

In this paper we present a new method for fast histogram computing. Based on the known tree-representation histogram of a region, also called reference histogram,, we want to compute the one of another region. The idea consists in computing the spatial differences between these two regions and encode it to update the histogram. We never need to store complete histograms, except the reference image one (as a preprocessing step). We compare our approach with the well-known integral histogram, and obtain better results in terms of processing time while reducing the memory footprint. We show theoretically and with experimental results the superiority of our approach in many cases. Finally, we demonstrate the advantage of this method on a visual tracking application using a particle filter by improving its time computing.