PixelOCR is a simple Java-based optical character recognition program designed to be easily trained through the use of font packs. Although this project
was more of a personal test application, it remains fully functional and ready to start converting images into text.
1. Font packs - A font pack is essentially a folder with a bunch of black and white .png files which are used as a reference. When you download the PixelOCR zip file, attached is a pre-made font pack. Remember to keep the jar file in the same directory as the "Fonts" folder. All your packs should be contained in the "Fonts" folder.
2. Fonts - The fonts themselves should have dimensions that fit perfectly around the character excluding effects (e.g. shadows). Remember when creating fonts, the characters themselves must be of a pure black color (R = 0, G = 0, B = 0). The first letter of the font filename will represent the optical character (E.png = 'E', e_.png = 'e').
3. Maximum character spacing is used to generate words from characters detected in the image, measured in pixels. The distance between one character and another may vary, so be sure to use a higher number. The space between characters should never be equal to or greater than the space between words. With this said, there must be at least a single space in-between characters for PixelOCR to work.
Packaged in the PixelOCR zip not only contains a pre-made font pack, but an image it was designed read. The image ("login.png") was taken from a popular online video game. If you select this image to be read, the colors which you can test against are pure yellow (R=255,G=255,B=0) and pure white (R=255,G=225,B=255). This means that when you scan the same image twice using different colors, you'll get two different results.
Remember to extract the files!