Image processing system utilizing neural network for discrimination between text data and other image data

• US 5,608,819 A
• Filed: 05/18/1994
• Issued: 03/04/1997
• Est. Priority Date: 07/19/1993
• Status: Expired due to Fees

• Current Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Original Assignee: Matsushita Electric Industrial Company Limited (Panasonic Holdings Corporation)
• Inventors: Ikeuchi, Takenobu
• Primary Examiner(s): Boudreau, Leo
• Assistant Examiner(s): Tran, Phuoc
• Application Number:
  US08/245,264
• Time in Patent Office: 1,021 Days
• Field of Search: 382/9, 382/14, 382/15, 382/155, 382/156, 382/157, 382/158, 382/173, 382/176, 382/190, 382/270, 382/282, 382/288, 382/304, 358/453, 358/455, 358/456, 358/457, 358/462, 358/464, 358/467
• US Class Current: 382/156
• 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 image processing system, which operates on an input image data stream consisting of successive multi-level values, effecting a plurality of respectively different types of image data processing in accordance with a plurality of different categories of the input image data, utilizes a neural network to assign each datum to a specific category, with resultant output signals from the neural network being used to select the appropriate type of image data processing for that datum.

Claims

1. An image data processing system for operating on a stream of input image data including successive multi-level picture element values supplied in respective sample periods, for selectively processing said input image data in accordance with a plurality of predetermined categories of said input image data, the system comprising:

data array extraction means for extracting from said input image data stream, in each of said sample periods, a set of picture element values including an object picture element value, said set corresponding to an array of successively adjacent picture elements;

characteristic quantity derivation means for operating on said set of picture element values to derive a set of characteristic quantities, said characteristic quantities expressing respectively different parameters of said array of successively adjacent picture elements;

a plurality of image processing means each coupled to receive said image data stream, functioning in each of said sample periods to generate respective processed image data in response to said object picture element value;

neural network means which has been subjected to a predetermined learning procedure, responsive to said set of characteristic quantities for judging said object picture element value as being within a specific one of said categories, and for producing output data indicative of said specific category; and

data selector means controlled by said output data from said neural network means for selecting processed image data produced from one of said plurality of image processing means.

2. An image data processing system according to claim 1, wherein said characteristic quantity derivation means comprises means for calculating, as one of said characteristic quantities, an average value of said set of picture element values.

3. An image data processing system according to claim 1, wherein said characteristic quantity derivation means comprises means for calculating, as one of said characteristic quantities, an average value of a subset of said set of picture element values, said subset corresponding to a linearly successively adjacent group of said picture elements within said array.

4. An image data processing system according to claim 1, further comprising means for delaying said output data produced from said neural network means during said each sample period, to obtain delayed output data, and for inputting said delayed output data to said neural network means in an immediately succeeding sample period.

5. An image data processing system according to claim 1, wherein said image data processing system is further supplied with operating condition data relating to an apparatus operating condition which affects said judgement executed by said neural network, and wherein said operating condition data are inputted to said neural network in conjunction with said set of characteristic quantities.

6. An image data processing system according to claim 1, wherein one of said plurality of image processing means comprises bi-level conversion means for converting said input image data to binary data by comparing said input image data with a variable-level threshold value, and wherein said variable-level threshold value is inputted to said neural network means in conjunction with said set of characteristic quantities.

7. An image data processing system according to claim 1 wherein one of said categories is a signal noise category and wherein said processed image data are binary data, and further comprising means (15) responsive to output data from said neural network indicating that said object picture element is within said signal noise category, for holding a processed image data output terminal (7) of said image data processing system at a fixed logic level.

8. An image data processing system according to claim 1, wherein one of said plurality of image processing means comprises bi-level conversion means for converting said input image data to binary data by comparing said input image data with a variable-level threshold value, and wherein said variable-level threshold value is inputted to said neural network means in conjunction with said set of picture element values.

9. An image data processing system for operating on a stream of input image data including successive multi-level picture element values supplied in respective sample periods, for selectively processing said input image data in accordance with a plurality of predetermined categories of said input image data, the system comprising:

data array extraction means for extracting from said input image data stream, in each of said sample periods, a set of picture element values including an object picture element value, said set corresponding to an array of successively adjacent picture elements;

a plurality of image processing means each coupled to receive said input image data stream, functioning in each of said sample periods to generate respective processed image data in response to said object picture element value;

neural network means which has been subjected to a predetermined learning procedure, responsive to said set of picture element values for judging said object picture element value as being within a specific one of said categories, and for producing output data indicative of said specific category; and

data selector means controlled by said output data from said neural network means for selecting processed image data produced from one of said plurality of image processing means, and for outputting said selected processed image data,

further comprising means for delaying said output data produced from said neural network means during said each sample period, to obtain delayed output data, and for inputting said delayed output data to said neural network means in an immediately succeeding sample period.

10. An image data processing system for operating on a stream of input image data including successive multi-level picture element values supplied in respective sample periods, for selectively processing said input image data in accordance with a plurality of predetermined categories of said input image data, the system comprising:

data array extraction means for extracting from said input image data stream, in each of said sample periods, a set of picture element values including an object picture element value, said set corresponding to an array of successively adjacent picture elements;

a plurality of image processing means each coupled to receive said input image data stream, functioning in each of said sample periods to generate respective processed image data in response to said object picture element value;

neural network means which has been subjected to a predetermined learning procedure, responsive to said set of picture element values for judging said object picture element value as being within a specific one of said categories, and for producing output data indicative of said specific category; and

data selector means controlled by said output data from said neural network means for selecting processed image data produced from one of said plurality of image processing means, and for outputting said selected processed image data,

wherein one of said plurality of image processing means comprises bi-level conversion means for converting said input image data to binary data by comparing said input image data with a variable-level threshold value, and wherein said variable-level threshold value is inputted to said neural network means in conjunction with said set of picture element values.

11. An image data processing system for operating on a stream of input image data including successive multi-level picture element values supplied in respective sample periods, for selectively processing said input image data in accordance with a plurality of predetermined categories of said input image data, the system comprising:

data array extraction means for extracting from said input image data stream, in each of said sample periods, a set of picture element values including an object picture element value, said set corresponding to an array of successively adjacent picture elements;

a plurality of image processing means each coupled to receive said input image data stream, functioning in each of said sample periods to generate respective processed image data in response to said object picture element value;

neural network means which has been subjected to a predetermined learning procedure, responsive to said set of picture element values for judging said object picture element value as being within a specific one of said categories, and for producing output data indicative of said specific category; and

data selector means controlled by said output data from said neural network means for selecting processed image data produced from one of said plurality of image processing means, and for outputting said selected processed image data,

wherein one of said categories is a signal noise category and wherein said processed image data are binary data, and further comprising means (15) responsive to output data from said neural network indicating that said object picture element is within said signal noise category, for holding a processed image data output terminal (7) of said image data processing system at a fixed logic level.

12. An image data processing system for operating on a stream of input image data representing an image and including successive multi-level picture element values supplied in respective sample periods and representing successive portions of said image to be transmitted, for selectively processing said input image data in accordance with at least first and second predetermined categories of said input image data, said first category identifying data expressing a picture region of an image and said second category identifying data expressing a region of printed characters of an image, the system comprising:

data array extraction means for extracting from said input image data stream, in each said sample period, a respective set of picture element values including a respective object picture element value, said respective set of picture element values corresponding to an array of successively adjacent picture elements;

characteristic quantity derivation means for operating on said set of picture element values to derive by arithmetic calculations, in each said sample period, a respective set of characteristic quantities expressing respectively different parameters of said array of successively adjacent picture elements;

a plurality of image processing means each coupled to receive said input image data stream, functioning in said each sample period to generate respective processed image data in response to said respective object picture element value;

neural network means which has been subjected to a predetermined learning procedure, responsive in each said sample period to said respective set of characteristic quantities for judging each said object picture element value as being within a respective one of said first and second categories and for producing, in each said sample period, output data indicative of said respective one of said categories; and

data selector means controlled by said output data from said neural network means for selecting, in each said sample period, processed image data produced from a respectively corresponding one of said plurality of image processing means corresponding to said respective category.