Forms such as business forms used in banks and post offices are
automatically classified using a form search apparatus and method. The
method of classifying forms comprises extracting features from the image
data of the input form and comparing the extracted features with stored
features of a set of template forms corresponding to a set of known
classifications of forms. The comparing step compares extracted features
which comprise attributes of tables contained in the template forms and
the input form respectively. The attributes of tables may be the number
of tables in the form, or the number of cells comprising the tables. An
approximate matching step is used to reduce the number of candidate
template forms.