Provided are techniques for identifying and extracting a sub-image in a document image file. Colors in a document image file are modified to form a modified document image file, wherein the document image file contains a first color, a second color, and a third color, wherein a threshold is used to determine whether each of different levels of the third color is to be one of the first color and the second color. Solid horizontal lines and solid vertical lines having one of a pre-defined width and a pre-defined height are removed from the modified document image file. A sub-image in the modified document image file is identified based on remaining solid horizontal lines and remaining solid vertical lines. A segment that includes the sub-image is extracted. Post-processing is performed on the segment.