Document page segmentation in optical character recognition

A method for use in an optical character recognition process, the method for detecting image regions in an input comprising a de-skewed gray scale image of a document, the method comprising the steps of:

• defining a background of the image region by detecting text in the image;
• decreasing image resolution to filter text from the image;
• creating a variance image from the filtered image;
• determining a threshold value defining a maximal local variance from pixels in the detected text;
• and
• generating a classification image by performing pixel based classification based on the maximal background variance to identify background pixels and image region pixels.

A method for page segmentation in an optical character recognition process to detect one or more textual objects or image objects in an input de-skewed gray scale image, the method comprising the steps of:

• creating an edge space from the gray scale image;
• applying central line tracing to connected components identified in the edge space to generate one or more native line candidates from the connected components;
• classifying the native line candidates as textual lines or non-textual lines so as to detect textual objects in the image;
• determining, from pixels in the detected text, a threshold value defining a maximal local variance;
• and
• generating a classification image by performing pixel based classification based on the maximal background variance to identify background pixels and image region pixels so as to detect image objects in the image.