Method and apparatus for automatic recognition of image and text/graphics areas on a master

• US 4,411,015 A
• Filed: 05/19/1981
• Issued: 10/18/1983
• Est. Priority Date: 05/23/1980
• Status: Expired due to Fees

• Current Assignee: Siemens AG
• Original Assignee: Siemens AG
• Inventors: Fuchsberger, Hermann, Scherl, Wolfgang, Wahl, Friedrich, Abele, Ludwig
• Primary Examiner(s): Boudreau, Leo H.
• Application Number:
  US06/265,107
• Time in Patent Office: 882 Days
• Field of Search: 340/146.3 MA, 340/146.3 AG, 340/146.3 Y, 340/146.3 H, 340/146.3 R, 358/21 R, 358/160, 358/161, 358/166, 358/167, 358/280, 358/282, 358/284, 382/18, 382/51
• US Class Current: 382/171
• CPC Class Codes: G06V 30/413 Classification of content, e.g. text, photographs or tables
• CPC Class Codes: H04N 1/40062 Discrimination between different image types, e.g. two-tone, continuous tone

Abstract

An apparatus and method for automatic recognition of image and text/graphics areas on a master undertake an opto-electronic scanning of the master with analog values obtained by the scanning being sampled and converted into digital information which is stored in a data field which is divided into a number of windows having edge lengths selected to be greater than the spacing between two successive lines of text. A grayscale value histogram is generated for each window and is statistically analyzed for determining whether the area of the master represented by the window is predominantly a text/graphics area or an image area based on known statistical characteristics generally related to each type of area.

Claims

1. A method for automatic recognition of image and text/graphics areas on a master comprising the steps of:

opto-electronically scanning said master for generating analog signals corresponding to printed matter on said master;

converting said analog signals into corresponding digital information;

storing said digital information in a data field in a memory which is subdivided into a plurality of partially overlapping windows forming a grid co-extensive with said data field, said windows having edge lengths selected to be greater than the known spacing between two successive lines of text on said master;

generating a grayscale value histogram for each window;

evaluating the maximum occurring brightness value of the entire master;

defining a lower brightness limit by a certain percentage of the maximum occurring brightness value in determining the percentage of samples within each grayscale value histogram which have a brightness between said maximum occurring brightness value and said lower brightness limit;

comparing said percentage of samples for each window to a selected decision threshold value;

classifying a window as an image area if said percentage of samples is less than said decision threshold value; and

classifying said window as a test/graphics area if said percentage of samples is equal to or greater than said decision threshold value.

2. The method of claim 1 wherein said selected decision threshold value is defined by the steps of:

determining a value S for each window according to the equation S=Σ

(1-μ

)³ h(i)/σ

³ ;

calculating a value C for the same window according to the equation C=Σ

(i-μ

)⁴ h(i)/σ

⁴,

wherein

i is the respective brightness value in the plurality of brightness values in the histogram for said window, h(i) is the number of samples having the brightness i in said window, μ

is the mean value of the overall information in said histogram, and σ

is the standard deviation for said histogram; and

wherein the step of classifying said window as a text area is defined by the step of;

classifying said window as a text area when said values S and C for said window are within a range on a surface having a vertical axis representing different C values and a horizontal axis representing different S values, said range defined by the relationships |C-C|<

b₁ C along said vertical axis and S<

b₂ along said horizontal axis, wherein C is a quadratic curve of the form C=a₂ S² +a₁ S°

a₀ wherein a₀, a₁, a₂, b₁ and b₂ are empirically determined constants.

3. An apparatus for automatic recognition of image and text/graphics areas on a master comprising:

a video scanner for opto-electronically scanning a master containing information to be calssified and stored as an image area or a text area, said video scanner generating analog signals corresponding to information on said master;

an analog to digital converter post-connected to said video scanner for converting said analog signals from said video scanner into digital information;

a semiconductor memory connected to said analog to digital converter for storing said digital information in the form of a plurality of overlapping windows forming a grid and each having edge lengths which are greater than the known spacing between two successive lines of text on said master; and

a microcomputer having access to said memory for generating a grayscale value histogram for each window, evaluating the maximum occurring brightness value of the entire master, defining a lower brightness limit by calculating a percentage of the maximum occurring brightness value and determining the percentage of samples within each grayscale value histogram having a brightness between said maximum occurring brightness value and said lower brightness limit, comparing said percentage of samples for each window to a selected decision threshold value, and for classifying the information in each of said windows as a text/graphics area if said percentage of samples for a window is equal to or greater than said decision threshold value and for classifying the information in a window as an image area if said percentage of samples is less than said decision threshold value.