Computer implemented method for segmenting a binarized document

A computer implemented method for segmenting a binarized document, the method comprising the steps of:

• extracting a plurality of connected components from the binarized document;
• and
• discriminating for at least one connected component of the plurality of connected components whether the at least one connected component of the plurality of connected components is a regular component based on a homogeneity level value,
• wherein the homogeneity level value is representative of a level of homogeneity within a local region, wherein the local region comprises the at least one connected component and at least one further connected component adjacent to the at least one connected component,
• wherein the homogeneity level value is determined based on a probabilistic cue integration of at least one first value representative of at least one image characteristic parameter determined for the at least one connected component and on at least one second value representative of the at least one image characteristic parameter determined for the at least one further connected component adjacent to the at least one connected component,
• wherein the probabilistic cue integration is based on at least one prior probability characterizing a cue probability.

A computer implemented method for segmenting a binarized document, the method comprising the steps of:

• extracting a plurality of connected components from the binarized document;
• and
• discriminating for at least one connected component of the plurality of connected components whether the at least one connected component of the plurality of connected components is a regular component based on a homogeneity level value,
• wherein the homogeneity level value is representative of a level of homogeneity within a local region, wherein the local region comprises the at least one connected component and at least one further connected component adjacent to the at least one connected component,
• wherein the homogeneity level value is determined based on a probabilistic cue integration of at least one first value representative of at least one image characteristic parameter determined for the at least one connected component and on at least one second value representative of the at least one image characteristic parameter determined for the at least one further connected component adjacent to the at least one connected component,
• wherein the probabilistic cue integration is based on at least one cue-specific distribution characterizing a cue-specific probabilistic homogeneity.