초록 ▼

A non-linear digital filter system for image data or other signal data suppresses undesirable noise while enhancing significant characteristics of the data. Data elements in a neighborhood (window) defined around and including an input data element are classified into characteristic classes based on

A non-linear digital filter system for image data or other signal data suppresses undesirable noise while enhancing significant characteristics of the data. Data elements in a neighborhood (window) defined around and including an input data element are classified into characteristic classes based on their parameter value range. The class with the maximum population is selected as most representative of the neighborhood. A data value associated with that class is chosen as the output of the filter for the input data element around which the neighborhood is defined, to thereby form a plurality filter. As an alternative, the class holding more than half the data elements in the neighborhood is chosen as most representative, to form a majority filter. In case there is no class containing more than half the elements in a neighborhood, the output of the filter equals the value characteristic of the class in which the input pixel itself falls. In a further heterogeneity filter version, a data element for whose neighborhood there is no majority class is replaced by a “1”; a pixel for whose neighborhood there is a majority class is replaced by a “0”. In the case of image data, the “1”value may represent a black intensity or a white intensity, in which case the “0”represents the opposite intensity.

대표청구항 ▼

In a method for processing a sequence of data elements each having a parameter value, the totality of such values being within a pre-determined range of values, the steps comprising (a) defining a pre-determined set of classes of parameter values, where each class corresponds to a sub-range of said

In a method for processing a sequence of data elements each having a parameter value, the totality of such values being within a pre-determined range of values, the steps comprising (a) defining a pre-determined set of classes of parameter values, where each class corresponds to a sub-range of said range of parameter values, and where each parameter value lies within a respective sub-range corresponding to one such class, (b) for each individual data element of said sequence, selecting a pre-determined multiplicity of neighboring data elements, (c) determining the sub-range for the parameter value of said individual data element and for the parameter value of each of said neighboring elements and increasing a count corresponding to the respective pre-determined class associated with that determined sub-range, (d) selecting without regard to rank order of parameter value a single one of said classes whose count is not exceeded by the count of any other class, and (e) replacing the parameter value of said individual element with a replacement value characteristic of said selected parameter value class.