Image binarizing apparatus

• US 5,617,484 A
• Filed: 06/06/1995
• Issued: 04/01/1997
• Est. Priority Date: 09/25/1992
• Status: Expired due to Fees

• Current Assignee: Olympus Optical Corporation Limited (Olympus Corporation)
• Original Assignee: Olympus Optical Corporation Limited (Olympus Corporation)
• Inventors: Wada, Toshiaki, Wang, Kangda
• Primary Examiner(s): Mancuso, Joseph
• Application Number:
  US08/466,362
• Time in Patent Office: 665 Days
• Field of Search: 382/156, 382/159, 382/161, 382/270, 382/271, 382/272, 382/273, 382/168, 382/172, 382/21, 382/23
• US Class Current: 382/172
• CPC Class Codes: G06V 30/10 Character recognition
• CPC Class Codes: G06V 30/162 Quantising the image signal
• CPC Class Codes: H04N 1/403 Discrimination between the two tones in the picture signal of a two-tone original

Abstract

An image binarizing apparatus, which comprises an image sensor for inputting a character image or a line image, an A/D converter for digitizing the output of the image sensor, a frame memory for temporary storage of the digital image, a window circuit for generating address information for dividing the stored image into a predetermined number of partial images, a brightness extractor for obtaining the highest brightness, lowest brightness and average brightness for the pixels for each block, a neural network which provides the optimum threshold value based on pre-learned data when receiving the highest brightness, lowest brightness and average brightness, a binarizer for binarizing each pixel of a partial image block from the frame memory based on the optimum threshold value ("white pixel" when the value of the brightness of the pixel is larger than the output value of the neural network, and "black pixel" when the value of the brightness of the pixel is smaller than the output value of the neural network), and another frame memory for storing the binarized image at a predetermined address.

Claims

1. An image binarizing apparatus comprising:

photoelectric converting means for photoelectrically converting an image into an electric image signal;

image signal separating means for separating an image signal acquired from said photoelectric converting means into a predetermined number of portions to yield partial images;

extracting means for extracting no more than three levels of brightness of a highest brightness, a lowest brightness and an average brightness as parameters from said partial images for a partial image block of interest and for each of a plurality of partial image blocks adjacent to said partial image block of interest;

threshold value computing means for inputting said extracted parameters comprised of no more than said three levels of brightness which are extracted from said partial image block of interest and all of said adjacent partial image blocks to a neural network, and causing said neural network to compute a single threshold value based on only said extracted parameters corresponding to said no more than three levels of brightness; and

binarizing means for selectively binarizing a plurality of pixels which form part of all of the pixels constituting an image, based on said threshold value acquired by said threshold value computing means.

2. The image binarizing apparatus according to claim 1, wherein said threshold value computing means includes:

another threshold value computing means for acquiring a temporary threshold value based on at least one type of parameter among said parameters extracted by a second extracting means;

compensation value computing means for inputting said extracted parameters to said neural network, and causing said neural network to compute a compensation value for a threshold value with one processing; and

compensating means for compensating said temporary threshold value with said compensation value.

3. The image binarizing apparatus according to claim 2, wherein said neural network learns by using as learning data a highest brightness, a lowest brightness and an average brightness which are extracted from each partial image selected from at least one sample character image, and by using as teaching data an optimum threshold value for the selected partial images or said compensation value for said temporary threshold value.

4. The image binarizing apparatus according to claim 1, wherein said threshold value computing means includes:

another threshold value computing means for acquiring a temporary threshold value based on at least one type of parameter among said parameters extracted by a second extracting means;

compensation value computing means for inputting said extracted parameters to said neural network, and causing said neural network to compute a compensation value for a threshold value with one processing; and

compensating means for compensating said temporary threshold value with said compensation value.

5. The image binarizing apparatus according to claim 4, wherein said neural network learns by using as learning data a highest brightness, a lowest brightness and an average brightness which are extracted from each of partial image blocks selected from at least one sample character image, and by using as teaching data an optimum threshold value for the selected partial image blocks or said compensation value for said temporary threshold value.

6. The image binarizing apparatus according to claim 1, wherein said neural network learns by using as learning data a highest brightness, a lowest brightness and an average brightness which are extracted from each of partial image blocks selected from at least one sample character image, and by using as teaching data an optimum threshold value for the selected partial image blocks.

7. An image binarizing apparatus comprising:

photoelectric converting means for photoelectrically converting an image into an electric image signal;

image signal separating means for separating an image signal acquired from said photoelectric converting means into a predetermined number of portions to yield partial images;

block selecting means for selecting a partial image block of interest from yielded partial images;

threshold value computing means for computing a plurality of threshold values for said partial image block of interest selected by said block selecting means;

extracting means for extracting no more than three levels of brightness of a highest brightness, a lowest brightness and an average brightness as parameters from said partial images for said partial image block of interest and for each of said partial image blocks adjacent to said partial image block of interest; and

threshold value selecting means for inputting said extracted parameters comprised of no more than said three levels of brightness which are extracted by said extracting means to a neural network, and causing said neural network to select only one of a plurality of threshold values computed by said threshold value computing means based on only said extracted parameters corresponding to said no more than three levels of brightness, and

wherein a plurality of pixels which form part of all the pixels constituting an image are selectively binarized on the basis of a selected threshold value.

8. The image binarizing apparatus according to claim 7, wherein:

said threshold value computing means computes for both white and black pixels in accordance with a statistical method a threshold value which is set in a range between a brightness in white pixels and a brightness in black pixels; and

said threshold value computing means computes for either white or black pixels a threshold value which is lower than that of a lowest brightness in said partial image block of interest and a threshold value which is higher than that of a highest brightness in said partial image block of interest.

9. The image binarizing apparatus according to claim 8, wherein said neural network learns by using as learning data said no more than three levels of brightness which include a highest brightness, a lowest brightness and an average brightness which are extracted from each of partial image blocks selected from at least sample character image, and by using as teaching data, detected information as to whether the partial image blocks include both white and black pixels, only white pixels, or only black pixels.

10. The image binarizing apparatus according to claim 7, wherein said neural network learns by using as learning data said no more than three levels of brightness which include a highest brightness, a lowest brightness and an average brightness which are extracted from each of partial image blocks selected from at least one sample character image, and by using as teaching data, detected information as to whether the partial image blocks include both white and black pixels, only white pixels, or only black pixels.