| PRTools contents |
HISTM
W = HISTM(A,N)
W = A*HISTM([],N)
C = B*W
C = HISTM(B,X)
C = B*HISTM([],X)
| Input | |
| A | Dataset or datafile for defining histogram bins (training) |
| N | Scalar defining number of histogram bins (default 10) |
| B | Dataset or datafile to be transformed into a histogram with predifined bins. |
| X | User defined histogram bins (centers) |
| Output | |
| C | Dataset or datafile with histogram bin frequencies |
For every object in the dataset B the set of feature values is mapped into a histogram, specifying for each bin the number of features having a value as specified for that bin. This is particular useful if the objects are images and the features are pixels. In that case for every image a histogram is found.
The dataset A may be used to find the proper histogram bins. In that case histograms with N bins are constructed between the minimum and maximum values over all objects in A.
Formally HISTM([],N) is an untrained mapping, to be trained by A as the dataset (datafile) A is used to determine the histogram bin centers. In case the bins are given like in HISTM(B,X) then we have a trained mapping. Consequently, if A is a datafile then in C = A*HISTM(A,10) all objects in A are processed twice. Once for determining the bin positions and once for filling them. If appropriate a command like C = A*HISTM(A(1,:),10) is thereby significantly faster, as it determines the bin positions by just a single object.
datasets, datafiles, mappings, hist,
| PRTools contents |