JPS59270A - Picture detecting system - Google Patents

Abstract

PURPOSE:To simplify a hardware, and to improve a picture at a high speed, by providing a detecting system which sets plural threshold levels to a dither matrix, and a detecting system which sets a single threshold level, and deciding by a logical operation whether a picture data is a shading picture or a character picture. CONSTITUTION:Whether each block is a shading picture or a character picture is decided from two deciding conditions of (1) whether the maximum level value of a variation portion of a tone between plural blocks of a dither picture exceeds a prescribed level or not, and (2) a picture element of a block of a single threshold level is not all white nor all black. A signal 1 in which the shading picture and the character picture are mixed is inputted in parallel to comparators 10, 11. The comparator 10 binary-codes the signal 1 by a single threshold level 2, and the comparator 11 binary-codes it by a dither threshold level 3. The binary-coded signal is brought to a logical operation and is inputted to a multiplexer 33, in which in accordance with the conditions (1), (2), it is made to pass through a simple binary data and a binary dither data when the input is a character picture, and when it is a shading picture, respectively.

Description

[Detailed Description of the Invention] U) Technical Field of the Invention The present invention employs systematic dithering by manually processing image data in which 21 direct drawings (abbreviated as character drawing 1 elephant) such as gradation painters and character 1 figures are mixed. 4 meetings detected using the method, 1 character image 1 elephant! ! I
This invention relates to a single-image detection method that improves I-sensitivity.

(2) Four points between the conventional technology and the conventional dither method, which expresses intermediate values in black and white binary values, has been used because of its gradation reproducibility and easy hardware configuration. It will be done. In the systematic dithering method, the intermediate A is inlaid by the density of the black l[III element, so it becomes -light +*
Good planning can be obtained for this.

But minutes! Due to the deterioration of the S ability, the image quality of the character strokes becomes worse in the 4th case where the Ittan Gaki and the character strokes II are mixed in the original A.

Figure 1 (→ and (b) show examples of one character stroke using the single-open dither method and the dither method for ■kanji and ■alphabet.As shown in the figure, when one straight stroke of 21 is detected using the dither method, It can be seen that because the boundaries between black and white images are not detected correctly, the fonts become distorted and the image quality deteriorates.

Conventionally, the method of improving the image quality in the order of character strokes by separating these two strokes and improving the image quality in the order of character strokes was proposed. Technical report 4jiVol 8
1, A 133. IE81-57 "Study of pseudo-halftone reproduction considering character stroke one image" and (2) % 1977-24165 One stroke one image identification method-j
There is.

Method (1) uses 1lIt for each small area of 4×41#.
The difference between the maximum mass and the minimum 1 scale is determined using the 1d model, and depending on whether this difference is smaller than a predetermined 1 scale, the gradation image 1 or the character image IIa is supplemented. Method (2) is dithered image 1
If the elephant is made to a specific size, for example 8 x 17 pixels, the distribution of the ll11i prime number of each pixel's gradation level determined by the threshold value of the dither matrix is calculated from the shape of this distribution. We are determining whether or not it is a consideration. However, (1)
The method is the maximum directness of multisource I [ki1 break 1d, the minimum lli
A special attack that performs rock filling for each block to search for gold over all pixels, method (2) is 1m in dither utii+#
+ 1il with 4 gradation levels for each element! A compulsory task is to count i prime numbers.

As a result, the hardware configuration becomes large and the processing speed cannot be increased.

(3) Purpose of the Invention The purpose of this book @ Ming is to input all the design data containing a mixture of striped and light paintings and characters and characters, and to detect it using the systematic dither method.
An object of the present invention is to provide a self-abuse detection method that improves the image quality of text.

(4) Structure of the Invention In order to achieve the purpose, the image detection method of the present invention uses a single image detection method in which a grayscale image and a character image are mixed.
IliIi image data is transferred to the first detection system in which a plurality of 1-1 directs are set as a dither matrix, and a single 1i141 direct is set to f.
52 plus is stored in parallel in the second detection system, the above-mentioned dizama) IJ Lux or a part thereof is set as one block, and the first detection system reads the block of interest. The maximum value of the level change?The means for detecting the second detection system is a single block.
Means for detecting a level change is provided, and when a different level change is detected immediately after detection by the means of the CX second detection system, the maximum lump of the change of the first detection thread means belongs to one or more cycles, the block to be noticed The data is determined to be a character 1 figure design, and it is simply -
The output of the detection number 1d based on the threshold value 1 is used as a % sign.

(5) Embodiments of the Invention Before explaining the present invention in detail, the systematic dither method will be explained.

FIG. 42 shows an example of a dither matrix of the systematic dither method, in which the numerals in the squares of bamboo ridges j having four rows are 1. When dithering the multiple masochism No. 16, which is now digitized into 16 values, a threshold 1 shift is set up in the dither matrix, and a multiple 1lt
l[! Compare it with the corresponding block of I11 elephants. Compare each pixel (try) once with the 61-1 direct TIJ of the dither matrix and the corresponding dither matrix, Qx
y>T+” and the 11th sho lI! Black level “1” for 11 elements
and set it to the white level l+oN when QXF (TO).

This is a method of converting a single grayscale image into two.

In the present invention, a dither matrix detection system with a plurality of 1-1 shifts and a detection system with a single shift is provided, and when a grayscale image 1#I and character defects are mixed in a document, the dither matrix detection system is provided. It is determined whether tti++1 in the IJX line is a grayscale image or a character image in units of matrix or partial area (block).

When the block is determined to be masochistic, it is detected by the dither matrix detection system, and when the block is determined to be a character image, it is detected by the single-I detection system. By processing in this way, the image quality of the first half of Tachibana (α) as shown in FIG. 1(0) is improved compared to the case of FIG. Single 1! You can get the same image quality as I4.

Character drawing 11 here! It is characterized by a sudden change in intensity and a high frequency of sudden I*la degree changes.

In the present invention, the above feature is taken, and the next two 1 (from 1 pair not noted) Kyonema block is fully supplemented whether it is a -fl image or character 1.

(1) Is the maximum amount of gradation change between multiple blocks of dither masochism equal to or higher than the level?

(ll) The self-family of a single 1-row block is not all white or all black. There are four black and white changes.

When the above conditions (1) and (n) are satisfied, the block is determined to be a character/painting character.

Judgment conditions (1) and (II) can be obtained using only Logic 4 operations as follows.

Regarding condition (1), first of all, assuming that the rm iron of interest has a uniform degree in each block, then the maximum change in gradation level between blocks in a 1
lLD is obtained as follows.

Now, pixels in a plurality of blocks are expressed as 5rnn. Here, Smn is day 10" or black 012, m is the block number among multiple blocks (m = 1 +..., M), n
is the threshold+1 rm ti (n=1 v...
・, N).

Each of the key levels has a maximum value of 1 and a small value, and the value of the value in equation (1) is taken as the sum of these four major values and the minimum value. Q expresses a discursive logical chaos.

In fact, V, the number of times in each block is not strictly uniform, but it is averaged across blocks, so the number of blocks k
There is no problem if you get 4 + hardness.

For the condition (If), each block has the following (Z) +
When equation (3) holds true, all pixels within the block become white and black, respectively.

(J 8mn=0 (m=L 2*””M)
12) n side 1 A Smn = 0 (m=1+ 2r=M)
"",! The Gs diagram is a configuration explanatory diagram of an embodiment of the present invention according to the above-mentioned principle, and shows a self-weight detection circuit.

In the same figure, in order to simplify the hardware configuration,
The size of one block is set to 2 x 4 pixels. Tsum #) 54
The top two rows of dither matrix C shown in Figure 2.

Take each F2 line as one block. Further, the numeric value of the multiplier nolock that determines the maximum weight of the change in the ISL tone is set to 4.

Comparator (COMP)10. Input it in parallel.

The comparator 10 converts the multiple self-abuse number 1d ■ into single-I flattery ■ 2
Become straight. The comparator 11 converts the multi-direction signal 1d into a single signal by dithering the signal into a single signal. Line counter (CNT) 15 and stroke JA number counter (CNT) 14
are read-only memory (ROM) with two pits each.
) 12 and causes 140M12 to generate a dither threshold sound as shown in FIG.

The simple 21 direct signal (2) which has been converted into a Z-C121 signal and the 2J direct dither number 100 (2) are input to shift registers (SR) 15 and 16, respectively, and are subjected to serial/parallel conversion. 4 bit parallel &
2 simple data stored in memory (MEM) 17
The 21st shift dither data (2) is stored in the memory (MEM) 19. Then, when the above operation is returned to C and the data of the 1st rhino is left in memories 17 and 19, respectively, the data of the 2nd rhino and the data of ■ and ■t are memorized respectively!
j (tvigM) Accumulate on 18 and 20.

Memories 17, 18, 19, and 20 have a configuration of 4 bits/word, and have 4 violets that can store exactly one line of data together with the valley memory.

The configuration and operation related to F-1 (one foot condition (1), (n)) will be explained below.

[1] When the determination condition (1) is completed, first the memory 1 is
Read t0 data for 4 words from 9 and 20 ■ Gold 1 word r
I will continue to do the following quickly in my heart.

Counters ((,'NT) 21 and 22 are memory 17.1B
, 19°20 address color/rideruru.

The counter 22 counts 2 pits at the f position and is used to read out 4 words.

First, the registers (REG) 26 and 28 are set with the data of (FF) One word t is read from each of the memories 19 and 20, and 8 bits of the read data 2 are stored in the register 260.
Cooking shell (AND) with Uchihiro and AND time P625, and the contents of register 2 and OR circuit 27 Diura 4+1
j(OR)t- and hits registers 26 and 28, respectively.

In the same manner as above, count and amplify one color/22t at a time, sequentially output data t and 4 from memory 19.20, AND (JRt), and set in registers 26.28.

this! Repeat db creation 4 times. By performing AND and OR operations, the registers 26 and 28 determine the minimum value of 1 and the maximum value of the gradation levels between the four blocks. Next, register 26
By putting the self-answer of 28 into the 4 summation M (EOR) of 29 and taking EORk, the maximum il of summerization between 4 blocks! ■ is required. This) direct [has] ROM 50
is entered in the address of ROM50 has a maximum of 1 direct [phase]
The output signal @ outputs "1" when the value is greater than or equal to a predetermined value, and outputs "0" when the value is less than a predetermined value.

This number 1d @ is set to 7 rigging 70 ng (FF) 31.

(2) NIJj: Determination of condition 1) Next, after clearing the color/re 22 t-, the counter 22
t1 count Arabure, memory 17.18.19.2
Output one word d each starting from 0. Simple 211 continued from Memo v 17 and 18! Data ■ connects gate circuits 26 and 24 to NAND1 path 25 't' NA
ND k, 0, and OR circuit 24 take 0Rt-.

The output 1d No. 0 and ■ are inputted to the AND circuit 32.

Output No. 1d (■) becomes '1' when the simple 2-directions of the block are not all black, and output No. 16 Q becomes *'i' when the pixels are not all white. In other words, the number 4 of the gate When there is even one change in black and white, -t, that is, when it is a character image, both the Ryokawa power signals become "1".

On the other hand, output No. 16 [stock] of flip-flop 61 according to judgment condition (1) is also ANI) l! l! It is input to the l path 62.

The output number 1 (B) of the AND circuit 62 satisfies the above-mentioned judgment condition (
■If C(II) holds, the key becomes 0'', otherwise it becomes 1@.Tsumashi, No. 1g■ becomes 1''' when the block of interest is supplemented with one character image, and p1 member is indifferent. It becomes 0'' when it is judged as mi+ broken.The 4L pure 2I of the block of interest
Direct data ■ and 2-direct dither data ■ are both branched and sent to multiplexer (MPX) 3! It is man-powered by l. The multiplexer 66 uses No. 1d (1) when the block of interest is added to the character stroke 1 and the simple 2-direction data (1), when the block of interest is determined to be the motive image, (2) The direct dither data ■ are all passed through. In this way, the output of the multiplexer 55 is obtained as 2 tE 100 ■ with improved image quality. In addition, the control circuit 40 has the above 19 counters 13, 14, 21. .22, shift register 1
5, 16, register 26.28. flip flop 51
4! A timing system will be used.

When the judgment of judgment condition (II) is repeated for four blocks to improve the image quality of four blocks, the counter 21 is incremented by one and the counter 22 is cleared to prepare for the image quality improvement of the next four blocks. Thereafter, the above operations are repeated in the same manner to improve the image quality.

In this way, as shown in Figure 5g1 (O), one character stroke whose image quality has been improved by the present invention is naturally shown in Figure 5g1 (O).
α) becomes the same.

In real ml+su, the block size is τ2×41[! I
Although we took the I element, the same can be said for the 4 elements of 4×4 pixels as well. This method can also be applied to formats other than the dither matrix shown in Figure 2.
It's huge.

In addition, although the example shows a circuit configured with wired logic, the circuit configuration can be further simplified by using a bit slice microprocessor or the like to act as a gate circuit controller. can do.

(6) As described in detail, according to the present invention, f! ! Late threshold 1st shift
An iM output line set in the dither matrix and a coalfield line set with a single 14th shift are set up, and -charcoal drawing order and letters are set.

Discussion of planning data mixed with figure artist in block units 4
Since it is determined by calculation whether it is shading or masochism, the hardware configuration is simple, and m+1 correction can be performed very quickly.

[Brief explanation of drawings]

Figures 1 (α) to (6) are examples of character strokes made by the VC of the present invention, 5g2 Figure V, and 1 of the Diri' matrix.
Example, Figure 6 V Mamoto @ Ming's, lI! ! There is a configuration explanatory diagram C for the 1M example. In the diagram, 10 and 11 are comparison 4, L5.14.21
．． 22 is color/re, 12, 30 is ROM, 15.16
is a shift register, 17°18, 19.20 is a memory,
25.32 is AND1gljii! ), 24°27 is O
R circuit, 26 is NAND m path, 29 is BOR circuit +1'
M, 26.28 is a register, 61 is a flip-flop,
66 is a multiplexer, and 4o is a control circuit. Patent Attorney Fujitsu Limited Sub-Agent Patent Attorney 1) Shige Saka Figure 1 (a), (1)) Figure 2

Claims (1)

[Claims] For stroke 1 data in which gradation planning and character 2 figure planning are mixed, a multi-a threshold is set in the dither matrix, and fc first detection system and single-interval 11i [ are set to be set to 942 detection system. They are stored in parallel, a number of pixels of the dither matrix III or a part thereof are set as one block, and the block to be focused on is included in the first detection system, so the maximum value of the change in gradation level between the blocks is detected. A method for detecting a level change in a single block focusing on the second detection system.
1. is provided, and when the maximum value of the change in step 4 of the 51!1 detection system is equal to or greater than a predetermined value, and the detection value of the second detection system has a different level of graying, the previous block of interest is A picture 1a habituation method characterized by determining data as a character 2 graphic artist and outputting a detection signal based on a single threshold.