A good example of binarization is the conversion of printed illustrations to their digital binary representations. First a little background. Vintage printed illustrations were generally continuous, i.e. a picture wasn’t made up of thousands of “dots”, as used in halftone printing which only simulates continuous tone. Ink was used, which made the images continuous – this is similar to the case of analog film, or even photographs created via the photogravure process, producing the continuous tones of a photograph.
The assumption is that something printed using black ink on a light background would be trivial to binarize? Here is an example of a print from a vintage book.
Is the histogram bimodal? Kind-of. There is a huge peak which represents the background (paper), and a comparatively small bump which represents most of the actual printed portion of the image.
Where is the threshold value? Somewhere between the small bump and the large peak. Using Otsu’s algorithm the threshold value is determined to be 137, which is just a little past midway. Here is the resulting binary image.
Most of the details seem to have been retained.