What is the difference between a simple histogram and a histogram LBP (local binary pattern)? Can someone provide the intuition behind histogram LBP?
I have read the wikipedia article for LBP and decided that they are using the word "histogram" incorrectly. They should really be referring to "binning". The former does "binning" as a pre-processing step, but the etymology of histogram suggests that it is only appropriate to call a graph which depicts frequencies using vertical columns a histogram.
The LBP is a rudimentary unsupervised classifier. It uses quadrants to bin and calculates local frequency maxima using nearest neighbor. All of these processes could be modified using other more sophisticated ML techniques.
EDIT: It looks like images can be further processed and indeed represented with some kind of histogram like here. It basically serves to graphically represent the features of an image, presumably if it's adequately normalized the bars should represent the intensity/darkness of various cells in the binned graphic. Each 2-D grid member is mapped (in a serpentine fashion) to the 1-D "feature" axis and the corresponding frequency is displayed. I haven't worked with this much, but I imagine the steps necessary to adequately normalize pictures "in the wild" would be impossible.