OCR - Optical Character Recognition

OCR also known as optical character reader is a device that translates a written or printed form of text into a digital file (eg. text is converted to ASCII codes) that can be edited by the computer using a word processor or the like.

Working principle of device:

OCR comprises of two parts: 1) Optical Scanner that scans the printed/written text and 2) Software that converts the scanned image of the text into the digital file such as .docx. Efficient OCRs need both the hardware and the software but certain inexpensive options rely entirely upon software. Trying to read a plethora of handwriting makes us cringe too, what to say of the poor computers! Printed text, due to its definitive nature, is not so difficult to transform, though. The optical scanner can be a simple camera or it could be a professional scanner. The only purpose in life of a scanner is to produce image or pdf file of the text. But the pdf or image file in itself is meaningless unless a software converts it to the digital format. The software employs two techniques namely Feature detection and Pattern recognition.

Pattern recognition works by comparing a pre-stored image (pattern) of a character with that of the scanned image. If the comparison yields true, Bingo! The OCR has successfully converted the character into its corresponding ASCII (or other character set) code. Otherwise it moves on to check for other characters in its database. The downside to this technique is that the OCR can detect only one kind of pre-defined fonts that are identical in height, width, breadth and stroke. Of course, running into such a kind of a font is rare coincident in the real world.

Feature detection/feature extraction/intelligent character recognition (ICR) Feature detection solves the above-mentioned problem. It doesn't compare the entire character with the printed image. Instead it will decompose the target character into more fundamental components and then compare those individual components with those of the scanned image. And if majority of the components match, then the character has been recognized successfully. For instance, 'A' can be subdivided into 3 components. Two vertical lines inclined at certain angle meet at a point. A horizontal line about half-way down from the vertex (meeting point of vertical lines) connects the two vertical lines. A combination of these 3 disparate parts makes up capital 'A'. Therefore, OCR can detect capital 'A' no matter in which font it has been typed. This eliminates the need to stick to a single font. Thus, we can say that feature detection is more edgy and versatile in nature than pattern recognition. Most modern systems have now switched to feature detection.

However, in the grand scheme of things we must defer to the fact that the ingenious human minds outwit the cunning computers hands down, howsoever smart they may have gotten! They definitely face-plant against the mysteriously inscrutable hieroglyphics of the illuminati scholars!

Device installation: OCRs as stated above are available in the shades of hardware as well as software. Nitro PDF, Abby Fine Reader, Simple OCR, Free OCR etc. A very many of them with advanced capabilities are proprietary software while the cowed down versions of the same are available for free. All that is needed is to procure the appropriate license of the software and install it on the machine. In case of hardware based OCR, the Real-World text is first scanned by the scanner and the scanned image is fed into the software. This scanner isn’t always natively supported by the operating system. Therefore, we need to install the drivers/software that are shipped with the product. (the drivers/software can alternatively be downloaded from the manufacturer’s website). The scanner can be connected to the computer via USB port.

Manufacturer of device, its models and prices:

1.     HP – Scanjet 200 Flatbed (Rs. 4,499), Deskjet 2131 All-in-One Printer (Rs. 2,299), 3000S2 Scanjet Scanjet Sheet Feed scanner (Rs. 24,231), N1920 Document Flatbed (Rs. 1,80,000) etc

2.     Canon– Canonscan Lide 120 (Rs. 3,503), Canonscan Lide 220 (Rs. 5,355)

3.     Epson – Perfection V39 (Rs. 15,355), Scanner V370P (Rs. 8,054),

4.     Kodak –I940 Scanner (Rs. 24,000), I2420 Scanner Single (Rs. 48,900) etc.

5.      Panasonic –  KV KV-S1027C/US Document Scanner (Rs. 1,00,546), KV-S1065C Document Scanner (Rs. 1,38,079) etc

Standard Configuration of device: OCR can come in two forms: 1) hardware as a scanner and 2) software. Paper is fed into the scanner which produces a pdf or an image file of the printed material. This digital file is converted to text by a software that accompanies the scanner. Software based OCRs directly take pdf or image files as input and translate them to editable text files (usually they are exported to .docx)

Cost: Standard home scanners are cheap and can be bought at about Rs. 2000. If high resolution, high precision and high quality scanners are desired than the cost shoots up to Rs. 25,000. However, heavy duty, high-end enterprise ready scanners easily cost anything above Rs. 1,00,000 and going up to Rs. 1,50,000. These scanners have very good image resolution and scan a very large number of pages per minute (ppm). In addition to that the OCR software that comes with these scanners is highly sophisticated and can translate contrived texts to their digital counterparts.

  

Market share of different models (Standard companies only):

l  HP

l  Canon

l  Epson

l  Kodak

l  Panasonic