Gray Level Technologies
Most common image processing solutions are based on black-and white (B/W) input images. The scanner produces a B/W image for processing; however, using B/W input may give poor results when the application works with a variety of unknown forms, or with forms that have backgrounds or background patterns. For bank checks, parcels, envelopes, credit card vouchers, boarding tickets, and so forth, the solution is to scan images in gray level and process the gray level images.
Image Binarization
Image binarization converts an image of up to 256 gray levels to a black and white image. Frequently, binarization is used as a preprocessor before OCR. In fact, most OCR packages on the market work only on bi-level (black & white) images.
To overcome real world problems, adaptive methods for image binarization are used.
The following figure demonstrates the binarization.
Figure 2 - The right column presents the results of binarization for the image on the left
Gap Filling Using Gray Level Analysis
One of the most disturbing problems associated with Form Dropout is the appearance of gaps in the data after subtraction. This happens when filled-in data is written over template data. Processing images in gray level may help to prevent these gaps.
In the following image, the filled-in data is written on the template lines.
The following figure demonstrates the process. a) Original image. b) After binarization. c) After Form Dropout (note the missing segments). d) Gray level data is used to fill in the missing segments.
Figure 3 -Gap filling example
Dropout Compression for Gray Level Form Images
Similar to bi-level (black and white) images, Form Dropout can realize significant compression ratios for gray level images. Maximal compression can be realized by taking bi-level dropout results. For applications that require very high image quality, gray level results can be used, and still achieve a very high compression ratio.
The following is an image of a bank check. Its GIF representation size is 500 KB.
Figure 4 - Gray level image of check
Below is the image reconstructed from the template and gray dropout with bi-level results. The image file size needed is 2 KB (compression factor of 250).
Figure 5 - Check image after gray level drop out
Below we see the image reconstructed from the template and gray dropout with gray results. The image file size needed is 5 KB (compression factor of 100).
Figure 6 - Reconstructed check image