# US 8,150,160 B2 · Automatic Arabic text image optical character recognition method

> Agents: this data is available as JSON from the patentgrep API and MCP server — see https://patentgrep.com/llms.txt

- Number: US 8,150,160 B2
- Document ID: US8150160B2
- Status: Granted
- Kind: B2
- Application number: 12/382,916
- Filed: 2009-03-26
- Published: 2010-09-30
- Granted: 2012-04-03
- Earliest claimed: 2009-03-26
- Assignees: King Fahd University of Petroleum and Minerals
- Inventors: Husni A. Al-Muhtaseb; Sabri A. Mahmoud; Rami Qahwaji
- Web page: https://patentgrep.com/patent/US8150160B2

## CPC

- G06F 18/295: {Markov models or related models, e.g. semi-Markov models; Markov random fields; Networks embedding Markov models}
- G06F: ELECTRIC DIGITAL DATA PROCESSING (computer systems based on specific computational models G06N)

## Abstract

The automatic Arabic text image optical character recognition method includes training a text recognition system using Arabic printed text, using the produced models for classification of newly unseen Arabic scanned text, and generating the corresponding textual information. Scanned images of Arabic text and copies of minimal Arabic text are used in the training sessions. Each page is segmented into lines. Features of each line are extracted and input to Hidden Markov Model (HMM). All training data training features are used. HMM runs training algorithms to produce codebook and language models. In the classification stage new Arabic text is input in scanned form. Line segmentation where lines are extracted is passed through. In the feature stage, line features are extracted and input to the classification stage. In the classification stage the corresponding Arabic text is generated.

## Claims

### Claim 1

A computer-implemented automatic Arabic text image optical character recognition method, comprising the steps of:
accepting as input an optical image of Arabic text and a predetermined selection of minimal Arabic text, the selection of minimal Arabic text representing substantially all Arabic letter shape occurrences likely to be input as text for character recognition;
arranging all of the input Arabic text into a plurality of text image rows;
segmenting each of the Arabic text input rows into a hierarchical stack of six segments vertically;
performing feature extraction on the six segments of Arabic text for at least one of the input rows;
inputting results from the feature extraction step into a Hidden Markov Model (HMM);
storing results from the HMM to an HMM models database and to a language model database;
performing a statistical syntactical analysis on the at least one input row of Arabic text;
storing results of the statistical syntactical analysis to the language model database;
selecting an optimized HMM from the HMM models database based on the statistical results stored in the language model database;
inputting additional Arabic text; and
performing text recognition on the additional Arabic text utilizing the optimized HMM.

### Claim 2 (depends on 1)

The computer-implemented automatic Arabic text image optical character recognition method according to claim 1, wherein each of the Arabic text input rows has a height and the step of segmenting each of the Arabic text input rows includes the step of segmenting the stack segments into an upper segment and a lower segment, the upper and lower segments each having a height of ¼ of the height of the corresponding Arabic text input row.

### Claim 3 (depends on 2)

The computer-implemented automatic Arabic text image optical character recognition method according to claim 2, wherein the step of segmenting each of the Arabic text input rows further includes segmenting the stack segments into four intermediate segments between the upper segment and the lower segment, each of the intermediate segments having a height of ⅛ of the corresponding Arabic text input row.

### Claim 4 (depends on 3)

The computer-implemented automatic Arabic text image optical character recognition method according to claim 3, further comprising the step of defining a 13-dimensional feature vector using the features extracted in said step of performing feature extraction.

### Claim 5 (depends on 4)

The computer-implemented automatic Arabic text image optical character recognition method according to claim 4, further comprising the step of sequentially sliding a feature extraction window horizontally across the Arabic text input row, adjacent feature extraction windows having an overlap of one pixel, the steps of performing feature extraction and defining a 13-dimensional feature vector being repeated for each successive sequential feature extraction window, the feature extraction window encompassing the six stack segments vertically.

### Claim 6 (depends on 5)

The computer-implemented automatic Arabic text image optical character recognition method according to claim 5, wherein the step of performing feature extraction further comprises;
extracting a 1st feature equal to summing black pixels in the lower segment of the feature extraction window;
extracting a 2nd feature equal to summing black pixels in the first intermediate segment immediately above the lower segment in the feature extraction window;
extracting a 3rd feature equal to summing black pixels in the second intermediate segment above the lower segment in the feature extraction window;
extracting a 4th feature equal to summing black pixels in the third intermediate segment above the lower segment in the feature extraction window;
extracting a 5th feature equal to summing black pixels in the intermediate segment immediately below the upper segment in the feature extraction window;
extracting a 6th feature equal to summing black pixels in the upper segment in the feature extraction window;
extracting a 7th feature comprising summing the first and second features;
extracting an 8th feature comprising summing the third and fourth features;
extracting a 9th feature comprising summing the fifth and sixth features;
extracting a 10th feature comprising summing the seventh and third features;
extracting a 11th feature comprising summing the fourth and ninth features;
extracting a 12th feature comprising summing the second, eighth and fifth features; and
extracting a 13th feature comprising summing the tenth and eleventh features.

### Claim 7 (depends on 1)

The computer-implemented automatic Arabic text image optical character recognition method according to claim 1, further comprising the step of performing statistical analysis of the input Arabic text to compute the probabilities of Arabic letters correctly recognized.

### Claim 8 (depends on 7)

The computer-implemented automatic Arabic text image optical character recognition method according to claim 7, wherein the Arabic text statistical analysis step comprises analyzing bi-grams.

### Claim 9 (depends on 1)

The computer-implemented automatic Arabic text image optical character recognition method according to claim 1, further comprising the step of using separate HMM models for the different shapes of each Arabic character.

### Claim 10

A computer-implemented automatic Arabic text image optical character recognition method, comprising the steps of:
accepting as input an optical image of printed Arabic text;
arranging all of the input Arabic text into a plurality of text image rows;
segmenting each of the input Arabic text rows into a hierarchical stack of six segments vertically;
performing feature extraction on the six segments of input Arabic text for at least one of the input rows;
inputting results from the feature extraction step into an optimized Hidden Markov Model HMM; and
performing text recognition on the input Arabic text image utilizing the optimized HMM.

### Claim 11 (depends on 10)

The computer-implemented automatic Arabic text image optical character recognition method according to claim 10, wherein each of the input Arabic text rows has a height and the step of segmenting each of the input Arabic text rows includes the step of segmenting the stack segments into an upper segment and a lower segment, the upper and lower segments each having a height of ¼ of the height of the corresponding input Arabic text row.

### Claim 12 (depends on 11)

The computer-implemented automatic Arabic text image optical character recognition method according to claim 11, wherein the step of segmenting each of the input Arabic text rows further includes segmenting the stack segments into four intermediate segments between the upper segment and the lower segment, each of the intermediate segments having a height of ⅛ of the corresponding input Arabic text row.

### Claim 13 (depends on 12)

The computer-implemented automatic Arabic text image optical character recognition method according to claim 12, further comprising the step of defining a 13-dimensional feature vector using the features extracted in said step of performing feature extraction.

### Claim 14 (depends on 13)

The computer-implemented automatic Arabic text image optical character recognition method according to claim 13, further comprising the step of sequentially sliding a feature extraction window horizontally across the input Arabic text row, adjacent feature extraction windows having an overlap of one pixel, the steps of performing feature extraction and defining a 13-dimensional feature vector being repeated for each successive sequential feature extraction window, the feature extraction window encompassing the six stack segments vertically.

### Claim 15 (depends on 14)

The computer-implemented automatic Arabic text image optical character recognition method according to claim 14, wherein the step of performing feature extraction further comprises;
extracting a 1st feature equal to summing black pixels in the lower segment of the feature extraction window;
extracting a 2nd feature equal to summing black pixels in the first intermediate segment immediately above the lower segment in the feature extraction window;
extracting a 3rd feature equal to summing black pixels in the second intermediate segment above the lower segment in the feature extraction window;
extracting a 4th feature equal to summing black pixels in the third intermediate segment above the lower segment in the feature extraction window;
extracting a 5th feature equal to summing black pixels in the intermediate segment immediately below the upper segment in the feature extraction window;
extracting a 6th feature equal to summing black pixels in the upper segment in the feature extraction window;
extracting a 7th feature comprising summing the first and second features;
extracting an 8th feature comprising summing the third and fourth features;
extracting a 9th feature comprising summing the fifth and sixth features;
extracting a 10th feature comprising summing the seventh and third features;
extracting a 11th feature comprising summing the fourth and ninth features;
extracting a 12th feature comprising summing the second, eighth and fifth features; and
extracting a 13th feature comprising summing the tenth and eleventh features.

### Claim 16 (depends on 10)

The computer-implemented automatic Arabic text image optical character recognition method according to claim 10, further comprising the step of using separate HMM models for the different shapes of each Arabic character.