Image processing system with layout analysis and method of operation thereof

A method of operation of an image processing system comprising:

• generating a binarized image by applying a mask threshold to a source image;
• detecting character targets within the binarized image, the character targets covering contiguous portions of the binarized image;
• forming connected neighbors by grouping the character targets having bounding boxes with a horizontal overlap greater than a horizontal overlap threshold;
• forming a text unit by grouping the character targets of the connected neighbors, the character targets having a character vertical overlap greater than a character vertical overlap threshold, the character targets having a character stroke size within a character stroke size threshold of the character stroke size of another of the character targets, and the text unit a portion of the source image;
• calculating a text unit baseline angle for rotating the text unit to the horizontal;
• and
• detecting an output text of the text unit for display on a device, wherein forming the text unit includes expanding each of the bounding boxes of the character targets horizontally by an extra border distance, the extra border distance calculated by a value α multiplied by the character height plus a border offset.

An image processing system comprising:

• a non-transitory memory for storing:
• a source image;
• a binarized image generated from the source image;
• a connected components module for detecting character targets within the binarized image, the character targets covering contiguous portions of the binarized image;
• a text unit module, coupled to the connected components module, for forming connected neighbors by grouping the character targets having bounding boxes with a horizontal overlap greater than a horizontal overlap threshold, and for forming a text unit by grouping the character targets of the connected neighbors, the character targets having a character vertical overlap greater than a character vertical overlap threshold, each of the character targets having a character feature within a feature threshold, and the text unit a portion of the source image;
• an identify baseline module, coupled to the text unit module, for calculating a text unit baseline angle for rotating the text unit to the horizontal;
• and
• an optical character recognition module, coupled to the text unit module, for detecting an output text of the text unit for display on a device;
• and
• a processor for processing the connected components module, the text unit module, the identify baseline module and the optical character recognition module, wherein the text unit module is for expanding each of the bounding boxes of the character targets horizontally by an extra border distance, the extra border distance calculated by a value α multiplied by the character height plus a border offset.