JPS6365778A - Picture identification system - Google Patents

Abstract

PURPOSE:To identify a character picture and a dot picture or the like at a high speed by checking the distribution of a signal at a proper area around a noted picture after a DC and a low frequency component is eliminated by a filter. CONSTITUTION:A filter circuit 1 eliminates the DC component and the low frequency component in an input picture signal. The output of the circuit 1 is stored by a shift register 4 to be a consecutive signal string. The content of the register 4 is transferred to the shift register 5. The counter 8 counts the signal string stored in the register 5 while dividing them as positive and negative portions. The count of the counters 8, 9 is added by an adder 10 and divided by a divider 11. A decision circuit 12 decides whether a picture is a character, a line picture, a dot picture or a photographic picture based on the result of operation of the adder 10 and the divider 11.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an image identification method for identifying type 1 of an image in an image processing device, etc.
The present invention relates to a method for identifying whether it is a line image, a halftone image, or a photographic image.

[Conventional technology and its problems]

In an image processing device, it is desirable to identify the type of image and perform processing according to the type of image.

For example, it is desirable to perform binary processing on text/line images to increase contrast, and to perform dither processing on photographic images to improve halftone reproducibility.

As disclosed in Japanese Patent Application Laid-Open No. 58-3374, the conventional image identification method examines the density difference between a pixel of interest and its surrounding pixels, and determines whether it is a character or line image based on the magnitude of the difference.
The purpose was to check whether it was a photographic image. However, when halftone dot images are processed using this method, they are recognized as text, line images, or photographic images, so it is difficult to apply this method to halftone dot images.

In addition, there is a method of recognizing halftone images by examining the correlation of images, such as the public axis of JP-A-56-132061, but with this method, it is impossible to recognize photographic images, and calculation It has the disadvantage that it requires a large amount of memory, a large amount of memory, and takes time to process.

The present invention was devised to eliminate the drawbacks of the conventional methods described above, and its purpose is to recognize any of text/line images, halftone images, and photographic images with a simple configuration. do.

[Means for solving problems]

In order to achieve the above object, the image identification method of the present invention removes direct current and low frequency components from an image signal, then compares it with a level near 0 to obtain a comparative output, and obtains the comparative output in the region of interest. By detecting the density of the image, it is possible to identify whether the image of the region of interest is a character/line image, a halftone image, or a photographic image.

As a method for detecting the state of the density of the comparison output, there is a method of obtaining a positive side comparison output and a negative side comparison output and detecting the ratio of the densities of the two comparison outputs, or a method of detecting the magnitude of the density of the comparison output. 6 [Function] To specifically explain the image identification method according to the present invention, the first
As shown in the figure, after the image signal is supplied to the filter circuit 1 to remove direct current and low frequency components, the distribution state detection circuit 2
The distribution state of the signal is checked by the judgment circuit 3 based on the result.
Then, it is determined whether the image is a character/line image, a halftone image, or a photographic image.

Now, the filter circuit 1 receives an image signal S1 shown in FIG.
is supplied, its direct current and low frequency components are removed, so the output of the filter circuit 1 becomes as shown in FIG. In addition, in FIGS. 2 and 3, a shows an example of a character/line image, b shows an example of a halftone image, and C shows an example of a photographic image.

The output of the filter circuit 1 is sent to the distribution state detection circuit 2,
It is compared with the level near 0, and the comparison output on the positive side is shown in Figure 4A.
The comparison output on the negative side is as shown in FIG. 4B. That is, the frequency of appearance of comparison outputs in the region of interest, that is, the density, differs depending on the type of image.

Furthermore, the density of the comparison output differs between the positive side and the negative side.

Therefore, for example, by detecting the density of one comparison output, it is possible to detect the M class of an image. Further, by performing addition and division between the positive output and the negative output, it is possible to obtain an output according to the type of image.

〔Example〕

DETAILED DESCRIPTION OF THE INVENTION Hereinafter, features of the present invention will be specifically described based on examples with reference to the drawings.

FIG. 5 shows an embodiment of an image identification device to which the image identification method according to the present invention is applied.

In the figure, numeral 1 indicates a filter circuit for removing direct current and low frequency components, and the image signal S shown in FIG. 2 is supplied to the filter circuit 1 as n-bit parallel data.

The filter circuit 1 removes direct current and low frequency components from the input image signal, and a signal as shown in FIG. 3 is obtained. That is, for character/line images, one input pulse is converted into one positive pulse with relatively large amplitude and two negative pulses with relatively small amplitude, and for halftone dot images, the halftone periodic component is converted into one positive pulse with relatively large amplitude and two negative pulses with relatively small amplitude. extracted, and there is almost no output for photographic images. As this filter circuit l, for example, F I R (
A digital filter that operates in synchronization with a finite pulse response type (acyclic type) shift clock S can be used. Note that the configuration of the filter may be two-dimensional or one-dimensional. Further, the coefficients and the number of stages of the filter are determined so that the filter frequency characteristics have a DC gain of 0 and allow only signals in the middle and high ranges to pass.

The output of the filter circuit 1 is stored in a shift register 4 as a continuous signal string. The number m of stages of the shift register used here must be large enough to store the number of pixels that is several times the halftone period.

1 in shift register 4 in synchronization with shift clock S.
Each time a pixel is input, the signal string in the shift register 4 is simultaneously transferred to the shift register 5.

Note that the shift register 4 has a serial input and parallel output, and the shift register 5 has a parallel input and serial output, and both are nXm matrix-like shift registers.

The signals stored in the shift register 5 are output in series by a clock S which is twice as high as the number of stages of the shift register compared to the shift clock SS synchronized with the pixels. Therefore, before the next transfer from shift register 4 to shift register 5, all outputs of shift register 5 are transferred to comparator 6.
Transferred to 7.

The output of the shift register 5 is supplied to comparators 6 and 7, which output a reference signal S having a value slightly higher than 0 level or O level, and a reference signal S having a value slightly lower than 0 level, respectively.
Compare the size based on 6.

By this comparison, the output of the shift register 5 is
The signal is separated into a positive signal train shown in FIG. 4 and a negative signal train shown in FIG. 4B. As these comparators 6.7, for example,
A digital comparator or the like can be used.

The outputs of the comparators 6 and 7 are input as clocks to counters 8.9, and the counters 8 and 9 count the number of pulses of the positive signal train and the negative signal train, respectively.

Shift clock S3 for transferring the contents of shift register 4 to shift register 5. Therefore, by clearing the counter 8.9, the signal string stored in the shift register 5 can be counted separately into positive and negative sides. The counts of counters 8 and 9 are added by an adder 10 and divided by a divider 11. In this case, since there are generally more signals on the negative side, the number of counts on the negative side is divided by the number of counts on the positive side. However, even in the opposite case, exactly the same identification can be made by simply changing the judgment conditions.

Then, based on the calculation results of the adder 101 and the divider 11,
According to the determination conditions described below, the determination circuit 12 determines whether the image is a character/line image, a halftone image, or a photographic image, and outputs respective identification outputs St, Ss, and S4.

In the case of character/line images, as can be seen from Figure 4, the number of signals (count numbers) on the negative side is greater than on the positive side, so the division result of the divider 11 will be a large value. Judgment can be made by At this time, a character/line image identification output S2 is output.

In the case of a halftone image, since the number of signals on the positive side and the negative side are approximately equal, it can be determined that the result of division is a value close to 1 and that the result of addition by adder 10 is a large value. At this time, a halftone image identification output S is output.

Furthermore, in the case of a photographic image, since the number of signals on both the positive side and the negative side is quite small, it can be determined that the addition result of the adder 10 is 0 or a value close to O.

At this time, a photographic image identification output S4 is output.

In the embodiment shown in FIG. 5, shift register 4.
5, comparators 6 and 7, counters 8 and 9, adder 10 and divider 11 correspond to the distribution state detection circuit 2 in FIG.

Another embodiment of the invention is shown in FIG. Note that parts corresponding to those in the embodiment shown in FIG. 5 are given the same reference numerals.

In this embodiment, a low-pass filter 13 is used to obtain a monomer similar to the filter circuit l shown in FIG. At the same time, the original image signal S1 is delayed by the delay circuit 14 by the same amount of time as the delay by the low-pass filter 13 to match the phases of both signals, and the subtracter 1
By subtracting the output of the low-pass filter 13 from the delayed signal in accordance with 5, it is possible to obtain a signal from which DC and low frequency components in the input image signal have been removed. That is, the low-pass filter 13, the delay circuit 14, and the subtracter 15 can obtain a signal similar to the output of the filter circuit 1 in the embodiment shown in FIG.

In the embodiment shown in FIG. 5, the filter circuit 1 works to emphasize middle and high frequencies, so high frequency noise is emphasized.
Images may be difficult to identify. On the other hand, the sixth
In the example shown in the figure, the high range is not emphasized, so
This has the effect of being less affected by high-frequency noise. Note that as the low-pass filter 13 used here, for example, an FIR type digital filter can be used as in the embodiment shown in FIG.

The output of the subtracter 15 from which the direct current and low frequency components in the input image signal have been removed is converted to a fixed base value S? at or near 0 in the comparator 6. It is compared with and binarized.

For example, if a positive side signal is used as the output of the comparator 6, the comparison output will be as shown in FIG.
The density of the comparison output is medium, high, or low depending on the type of image, such as a line image, a halftone image, or a photographic image. As the output of the comparator 6, either a positive side signal or a negative side signal can be used depending on the comparison method. Although not shown in the figure, two comparators may be used to calculate the sum of both.

In other words, as can be seen from FIG. 4, it is a character/line image.

Regarding halftone images and photographic images, there is a large difference in density between the comparative outputs of these three, so there is no big difference in identification no matter which output is used (i. Any kind of comparison output is acceptable as long as it is possible.

The output signal string of the comparator 6 directly serves as a clock for the counter 8, and the counter 8 counts the signals. counter 8
is reset by a signal obtained by dividing the shift clock S of the low-pass filter 13 by the frequency divider 16, so that the count 8 counts the number of signals during this reset signal. Therefore, this frequency division may be set to a value about the same as the number of stages of the shift register in the embodiment of FIG. In this embodiment, unlike the embodiment shown in FIG. 5, a shift register is not used, so identification is made for each subregion at the reset/7 reset interval, but the circuit can be simplified.

The output of the counter 8 is compared with the first and second reference values 19.20 by comparison 817.18. The value of the first reference value 19 is set smaller than the count number in the case of a halftone image,
Set to a larger value than the count number for text/line images. However, in reality, it may not be possible to find an appropriate value depending on the image reading system or the condition of the document.
Set the value that gives the best recognition result. Further, the second reference value 20 is set to a value smaller than the count number in the case of a character/line image, in fact, a value close to 0.

The determination circuit 12 determines whether the image near the pixel of interest is a character/line image, a halftone image, or a photographic image based on the comparison results of the comparators 17 and 18. That is, counter 8
If the count number is smaller than the first reference value 19 and larger than the second reference value 20, it is determined to be a character/line image and output S2, and if it is larger than the first reference value 19, it is determined to be a halftone image and output. S, and if it is smaller than the second reference value 20, it is determined that it is a photographic image and an output S4 is output.

In the embodiment of FIG. 6, a low-pass filter is used to
Since the same effect as in the embodiment shown in the figure can be obtained, it is difficult to be influenced by noise and identification can be ensured.

Furthermore, as shown in FIG. 6, since there is no need to use a shift register, the circuit can be made into single-layered circuits, and an inexpensive device can be provided.

Furthermore, since the counter 8 is operated at a frequency lower than the clock frequency of the shift clock S, the overall processing speed can be increased, making it suitable for high-speed processing.

Note that the low-pass filter 13 in the embodiment shown in FIG.
．． The filter circuit composed of the delay circuit 14 and the subtracter 15 can be directly replaced with the filter circuit 1 of the embodiment shown in FIG.

Further, since the image is two-dimensional, it is desirable to process this embodiment two-dimensionally, but if you want to reduce the cost, it is also possible to perform identification using -dimensional processing. However, lines of text/line images in the same direction as the processing direction cannot be distinguished from photographs, so a method is proposed that allows different characters/pictures to be identified by the direction component, for example, Japanese Patent Laid-Open No. 58-3374.
Identification can be ensured by using the method shown in the above publication.

In the embodiment shown in FIG. 5, the shift register 4 is configured to directly store the signal from the filter circuit 1.
Exactly the same effect can be obtained by first converting the signal from the filter circuit 1 into a binary value using the comparators 6 and 7, and then storing the outputs of the comparators 6 and 7 separately in a 1-bit shift register.

Further, in this specification, the embodiments shown in FIGS. 5 and 6 are given as examples of the distribution state detection circuit 2 shown in FIG. 1, but the embodiments are not limited to these, and various methods may be adopted. obtain.

〔Effect of the invention〕

As described above, in the image identification method of the present invention,
After removing direct current and low frequency components using a filter, the signal distribution state in a suitable area around the image of interest is examined, and depending on the difference in the distribution state, it is determined whether it is a character/line image, a halftone image, or a photographic image. identify As a result, the above 31! ! It becomes possible to identify similar images. Also,
Since it is possible to identify three types of images using the same circuit, the circuit configuration can be simplified compared to those in which a dedicated identification circuit is provided for each image. Furthermore, since complicated processing such as image correlation detection is not required, high-speed, large-capacity memory is not required, and inexpensive and high-speed identification is possible.

[Brief explanation of drawings]

Fig. 1 is a configuration diagram showing the principle of the present invention, Fig. 2 is a waveform diagram showing an image signal, Fig. 3 is a waveform diagram showing a signal after filter processing of the image signal shown in Fig. 2, and Fig. 4A. is a waveform diagram showing the positive side comparison output, FIG. 4B is a waveform diagram showing the negative side comparison output,
FIG. 5 is a block diagram showing one embodiment of the invention, and FIG. 6 is a block diagram showing another embodiment of the invention. 1: Filter circuit 2: Distribution state) Expenditure circuit 3.12
: Judgment circuit 4.5: Shift register 6.7: Comparator 8, 9: Counter 10: Adder 1
1: Divider 13: Low-pass filter

Claims (1)

[Claims] 1. After removing direct current and low frequency components from the image signal, by comparing this with a level near 0 to obtain a comparison output, and detecting the density of the comparison output in the region of interest. . An image identification method characterized by identifying whether the image of the region of interest is a character/line image, a halftone image, or a photographic image. 2. The comparison output is a positive side comparison output and a negative side comparison output,
The image identification method according to claim 1, characterized in that the ratio of densities of both comparison outputs is detected. 3. The image identification method according to claim 1, wherein the density of the comparative output is detected.