JPS62171085A - Character image binarization device - Google Patents

Abstract

PURPOSE:To execute the stable binarization regardless of a concentration level change by obtaining the mean value of a concentration value, knowing a ground level in a local area and executing the binary decision with this when an ambient picture element is a ground picture element. CONSTITUTION:The image signal obtained by scanning a character image with an image input sensor 11 is quantized for respective picture elements in an A/D converter 12 and inputted and stored in a two-dimensional local memory 13. The value (variable density value of the picture element in recording media) quantized concerning a central picture element and plural ambient picture elements as the representative picture element to be representative of the local area are read, the binarization of the central picture element is executed by a binarization part 14 and the result is sent to a character recognizing part 15. The recognizing part 15 executes the character recognition from the pattern of the binary coded dot. Thus, the binarization of the attention central picture element is executed in accordance with the variable density mean value of the ground picture element in the vicinity of the attention central picture element and the stable binarization is executed.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention relates to a sensor (image input sensor) that scans a character image area to be read in an optical character reading device.
Regarding the image signal obtained from the image signal, the character used for binarization is "0" for the signal representing the ground part of the recording medium on which characters are recorded, and "1" for the signal representing the character pattern part. The present invention relates to an image binarization device.

[Conventional technology]

FIG. 4 is an explanatory diagram showing an example of an image to be read (object to be binarized) in the present invention.

In other words, it will be understood that the target is ordinary characters such as alphabets and numbers.

Fig. 4A is an explanatory diagram of the principle of binarization used in a conventional character image binarization device, and shows the shading level of the character image shown in Fig. 4 and the ground portion in pixel units. A gray histogram is shown in which the frequency of appearance (number of pixels) of the level is plotted on the axis and the vertical axis is plotted.

Conventionally, a histogram is taken in this way, the peak L1 of the shading level of the character part and the peak L2 of the shading level of the background are checked, and the minimum frequency shading level at the base of both peaks is calculated as 2.
Decide on the value conversion level S, and use the level S to convert the character image into 2
It was valued.

However, as can be seen from Figure 4A, there is a portion ΔH where the character shading frequency distribution range H1 and the ground gradation frequency distribution range H2 overlap each other, and in this ΔH portion, binarization becomes unstable. turn into. This is because the binarization level was determined from the histogram obtained for one entire screen, and the unevenness of the shading values of characters and background depending on the position within the screen was not taken into account.

[Problem that the invention seeks to solve]

Therefore, in the present invention, we aim to perform a stable binarization operation for the pattern part and the ground part of the image signal obtained from the sensor. The problem to be solved is to make this possible regardless of positional variations such as sensor sensitivity.Therefore, an object of the present invention is to provide a character image binarization device that makes the above possible. shall be.
[Means and effects for solving the problem] In order to achieve the above object, the present invention includes the following:
Utilizes two-dimensional local memory. This two-dimensional local memory pixelizes and quantizes a two-dimensional image signal obtained by capturing an image with a two-dimensional imaging device such as a television camera, and stores it two-dimensionally in the memory. It is possible to retrieve pixel data in any desired local area of the memory (as a two-dimensional local memory,
The present inventor has proposed a convenient method in which the possible local area can be changed freely in Japanese Patent Application No. 60-223697).

Therefore, in the present invention, a local area larger than the target character size is considered as a local area in a two-dimensional local memory, and a central pixel and a plurality of surrounding pixels in the area can be read out as representative surface elements. The purpose of this method is to identify ground pixels from among a plurality of surrounding pixels, and to average and binarize the gray values of these ground pixels.

That is, in the local area assumed in the present invention, if the central pixel is a pixel on the character pattern,
It is believed that many of the surrounding pixels are pixels on the background, that is, the area corresponding to the ground.

The present invention is particularly effective when applied to such character images.

〔Example〕

The present invention will now be described in detail with reference to the drawings.

FIG. 5 is an explanatory diagram showing an example of a character image to be binarized by the binarization apparatus according to the present invention, and also showing the waveform of an image signal obtained by scanning the image.

That is, the image signal to be binarized by the binarization apparatus according to the present invention is an image signal obtained by scanning general characters (alphabet, hiragana, etc.) as shown in FIG.

For example, the waveform of an image signal obtained by optically scanning a line x1 that overflows in the main scanning direction is like Sx, and the waveform of an image signal obtained by optically scanning a line y1 that collapses in the subscanning direction is Sx. The waveform of the image signal is similar to Sy, but the DC level on the line segments that make up the characters fluctuates greatly depending on the character position, and conventional binarization devices cannot achieve stable binarization. It's difficult. However, when considering a local region of the size of a character, it will be recognized that within that range, the variation in the gray level of the background is much smaller than that of the character portion.

FIG. 1 is a block diagram showing the overall configuration of an optical character reader (OCR) incorporating a binarization device according to the present invention.

In the figure, 11 is a sensor (image sensor) that optically scans a character image written on a recording medium such as paper and outputs an image signal. Specifically, it is a sensor that outputs an image signal by scanning a character image written on a recording medium such as paper. It can be thought of as equivalent to

12 is an A/D converter that quantizes the image signal from the sensor 11 and outputs it as a digital signal of multiple bits (for example, 6 bits, 64 gradations); 16 is the aforementioned two-dimensional local memory; and 14 is a binarization unit. , 15 is a character recognition section.

As mentioned above, the two-dimensional local memory 13 is explained in detail in the specification of Japanese Patent Application No. 60-223697. This will be explained with reference to the figures.

FIG. 2 is an explanatory diagram showing an example of a possible local area in the two-dimensional local memory 13 and a pixel arrangement representative of the area.

That is, a partial character image area (M dots) x (N dots) (3 dots x 3 dots in this example) is designated as a local area, and the center pixel therein is designated.

(represented by a black circle mark) and surrounding pixels D1 (represented by a white circle mark) that can be selected arbitrarily are considered as representative surface elements representing the local area, and the quantized values of these pixels are simultaneously read out. It is the two-dimensional local memory that makes it possible to store 8 images. Therefore, if you shift the center pixel one dot at a time along the scanning direction on the scanning screen, the local area centered on the center pixel will also move in the same way, so any pixel on the entire screen will be shifted one dot at a time along the scanning direction. When a pixel is used, surrounding pixels can be considered, and the quantized value of each pixel at that time can be known.

FIG. 3 is a block diagram showing details of the binarization section 14 (an embodiment of the present invention) in FIG. 1.

In the same figure, 601 to 608 are peripheral pixel determination units, 309 and 310 are adders, 511 is a divider, 612 is an absolute difference calculator, and 313 is a comparator.

Returning to FIG. 1, the overall circuit operation of the optical character reading device will be briefly described. An image signal obtained by scanning a character image by the image input sensor 11 is quantized for each pixel by the A/D converter 12, and input to and stored in the two-dimensional local memory 13. Next, from the two-dimensional local memory 13, the quantized values (shading values of pixels on the recording medium) of the central pixel as a representative surface pixel representing the local area and its surrounding pixels are read out, and the binarization unit At step 14, the central pixel is binarized and the result is sent to the character recognition section 15. The character recognition unit 15 performs character recognition of the target image scanned by the sensor 11 from the binary dot (pixel) pattern obtained as described above.

Next, the circuit operation of the binarization section 14 as an embodiment of the present invention will be explained with reference to FIG.

As already explained, the binarization section 14 has a configuration as shown in FIG. 3.

The human power signal p in FIG. -p8 is the output from the two-dimensional local memory 13 representing the local area, I)Q is the gray value signal of the center pixel of interest (the center pixel Do in FIG. 2), and p1 to p8 are the peripheral pixels ( 8 in Figure 2
This is the gray value signal of the surrounding pixels D1). α is the setting signal, that is, the setting human power indicating the minimum character signal amplitude,
β is a setting input indicating the binarization level, and is a value determined relative to the grayscale values of other pixels in the local area.

The ground peripheral pixel determination units 301 to 308 calculate the absolute difference between the X input (shading value of the center pixel of interest) and the X input (graining value of each peripheral pixel), and the difference is determined by the two manual inputs (the minimum character signal amplitude). When it is larger than the input (setting input as ), outputs the It is supposed to be done.

That is, the ground peripheral pixel determination units 301 to 308 select one center pixel p. and the eight surrounding pixels p1 to p8, and when the difference in density is larger than the minimum character signal amplitude, the surrounding pixels p1 to p8 are determined to be other pixels (not character pixels), and output. The gray value signals of other pixels are applied from the p terminal on the side to the adder 609 for addition, and on the other hand,
When the shading difference is smaller than the minimum character signal amplitude, K determines that the surrounding pixels p1 to p8 are character pixels, and in this case outputs zero from the output terminal p, so that the shading value signal of the character pixel is It is prevented from being input to the adder 409.

Further, the peripheral pixels p1 to p8 are determined by the ground peripheral pixel determination units 301 to 3.
08, when the pixel is determined to be another pixel as described above, a logic "1" is output from the Q terminal on the output side of each determination unit 601 to 308 and sent to the adder 310, and on the other hand, the pixel is determined to be a character pixel. Since the button is configured so that logic "0" is output from the Q terminal when the
By summing up the number of input logic "'1"'s, it is possible to know how many pixels among the surrounding pixels p1 to p8 are other pixels.

In this way, the adder 309 calculates and outputs the sum a of the gray values of the surrounding pixels determined to be other pixels.
10 calculates and outputs the number of peripheral pixels determined to be other pixels. Therefore, the divider 311 calculates the average gray value of other pixels among the surrounding pixels by executing the a/b calculation. The absolute difference calculator 312 calculates the center pixel p of interest. The absolute difference between the gradation value of and the average gradation value of other pixels from the divider 511 is calculated. This absolute difference is determined by the comparator 313
It is compared with the 2'-valued level β at , a binary determination is made, and a 2-valued output B can be output.

By doing this, the binarization of the center pixel of interest is
This can be carried out according to the gray level average value of other pixels near the center pixel of interest, and stable binarization can be performed.

〔Effect of the invention〕

According to this invention, the binarization of each pixel constituting a character image is replaced with the binarization of the center pixel of interest in a local area selected larger than the character size, and a plurality of peripheral pixels in the local area are referred to, When the surrounding pixels are other pixels, the ground level in the local area is determined by calculating the average value of the grayscale values, and this is used to perform binary judgment.
It has the advantage that 2#L conversion can be performed extremely stably, regardless of level fluctuations in the shading values of characters and background depending on the location within one screen.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the overall configuration of an optical character reading device (OCFt) incorporating a binarization device according to the present invention;
FIG. 2 is an explanatory diagram showing an example of a local area that can be taken in a two-dimensional local memory and a pixel arrangement representative of the area, and FIG. A block diagram showing details, FIG. 4 is an explanatory diagram showing an example of an image to be read, FIG. 4A is an explanatory diagram of the principle of binarization used in a conventional character image binarization device,
FIG. 5 is an explanatory diagram showing an example of a character image to be binarized by the binarization apparatus according to the present invention and the waveform of an image signal obtained by scanning the image. Explanation of symbols: 11... Image human power sensor, 12...
A/D converter, 16... Two-dimensional local memory, 1
4... Binarization unit, 15... Character recognition unit, 301 to 308... Curved area surrounding pixel determination unit, 309.5
10・Song・Adder, 311...Divider, 612
...Absolute difference calculator, 313... Comparator Fig. 1 Fig. 2 M dot Fig. 4 Fumiko's branching rod surrounding Fig. 5 0□ Pass

Claims (1)

[Claims] 1) A character image input sensor that decomposes a target character image area into pixels, scans in the main scanning direction and sub-scanning direction, and detects and outputs a character image signal for each pixel; an A/D (analog/digital) converter that inputs the character image signal, quantizes it, and outputs it; and an A/D (analog/digital) converter that receives the quantized signal from the converter and places the pixel of interest on the character image area as the center pixel; a two-dimensional local memory that sequentially outputs quantized signals for the central pixel and a plurality of surrounding pixels in a two-dimensional local area selected to have a predetermined size for each pixel of interest in synchronization with the scanning;
Selecting only the pixels representing the ground portion in the image area from among the peripheral pixels, determining the average value of their quantized signals, and calculating the absolute difference between the average value and the quantized signal of the central pixel. A character image 2 characterized in that it is equipped with a calculation means for calculating, and a binarization means for comparing the absolute difference with a certain set value and determining the binarization of the central pixel based on the magnitude relationship. Value device.