Binarizing an image

A method for binarizing an electronic image with text of different sizes, the method comprising:

• acquiring, by one or more processors, an electronic copy of the electronic image;
• performing, by one or more processors, a smoothing function on the electronic image;
• creating, by one or more processors, a gradient image from the electronic image;
• identifying, by one or more processors, objects in the gradient image by identifying boundaries of the objects;
• identifying, by one or more processors, likely-text objects from the identified objects;
• classifying, by one or more processors, likely-text objects as text objects of a first size or as text objects of a second size, wherein text of the second size is smaller in at least one aspect than text of the first size;
• identifying, by one or more processors, a region encompassing at least some of the likely-text objects of the second size;
• performing, by one or more processors, a binarizing function on said identified region that encompasses at least some of the likely-text objects of the second size, wherein the binarizing function includes:
• creating a binarization threshold map for the identified region using values from the (non-smoothed) electronic copy of the image;
• and
• comparing a brightness value of each pixel of the identified region of the electronic copy of the electronic image with a respective threshold binarization value from the binarization threshold map.

A device for binarizing and processing an electronic image, the device comprising:

• a processor;
• a memory in electronic communication with the processor, the memory configured with instructions for:
• creating a gradient image from the electronic image that includes text;
• identifying one or more objects in the gradient image;
• identifying one or more boundaries in the gradient image for each identified one or more objects based on a contrast value for the respective identified object;
• identifying any separation objects from the identified one or more objects in the gradient image;
• assigning a threshold binarization value for use with each pixel of the electronic image;
• storing said assigned threshold binarization values in a binarization threshold map, wherein the binarization threshold map includes a same threshold binarization value for each identified image object identified bin the gradient image, and wherein the assigned threshold binarization values are adjusted based in part upon any neighboring identified separation objects;
• and
• creating a binarized representation of the electronic image by comparing a brightness value of each pixel of the electronic image with a respective threshold binarization value from the binarization threshold map.