You are on page 1of 22

OCR

Optical Character Recognition

OCR
OCR stands for Optical character recognition

It is the process of translating scanned images of typewritten text into machineeditable information
Process involves analyzing the content and

recognizing the identity of any character contained in a digitized image and returning its information.

Block Diagram for OCR

PROCESSES INVOLVED
1.CLASSIFICATION a. Training Preprocessing Feature extraction Model estimation b.Recognition (testing) Preprocessing Feature extraction classification

PREPROCESSING
It involves Binarisation Morphological operators segmentation

BINARISATION
Converts a grey level TIFF (8 bit) image to

binary image.
Histogram based gloproachbal threshold

approach
Helps in segregating background from text.

MORPHOLOGICAL OPERATORS 1.ORIENTATION

ORIENTATION

ORIENTATION PROBLEMS

2. SKEW CORRECTION
Determining the degree of skewness

Use headline or page edges for correction by

rotating the image

SEGEMENTATION
Basic idea:- dissect the image of a line of text

into location between characters, i.e. , character breaks. Character break depends upon: Font Type sizes Printing quality

TYPE OF SEGEMENTATION
Bottom up

Top down

mixed

BOTTOM UP APPROACH
Segmentation starts with individual letters on a page, then based on text layout conventions, groups

letter into words, words into paragraphs, and so on

TOP DOWN APPROACH


Top down approach takes advantage of the fact that formatted documents have

margins surrounding each region. The page can be subdivided into regions depending on white spaces in the document.

PROBLEMS IN SEGMENTATION
OVER SEGMENTATION Dot matrix printing or insufficient inks Characters tend to be fragmental UNDER SEGMENTATION Ink smudging Small fonts Signatures

OVERCOMING SEGMENTATION PROBLEMS


Separation by Valley of Vertical Projection Searches for vertical white space between characters. Projection of character pixels along the vertical direction and detection of valleys.

Cut And Test


This technique dissects the character image at several candidate location s and evaluates the result of the segmented pieces. The candidate locations are determined by considering factors such as the average Break point distances.

BACKGROUND NOISE

Textual Noise

Non-Textual Noise

Extraneous symbols from Black Borders. Speckles. the neighboring page. Hand Written Material.

TEXTUAL NOISE

NON TEXTUAL NOISE

NOISE REMOVAL Page frame Rectangular region enclosing all the foreground pixels in the document image. Parameters : Size Aspect Ratio Limitations: Fail if characters from adjacent page are present.

CONCLUSION
OCR is useful in converting type written text into machine editable format which can be further processed as per requirements. Digitization of Libraries. Digitization of old manuscripts. Helpful for visually challenged .

You might also like