JPH08294002A - Image division method and image division device - Google Patents

Abstract

PURPOSE: To provide the image division method in which a block in a 2-color image area is discriminated as a 2-color image area independently of a size of a character and a graphic and an image is divided without limiting colors of a 2-color image. CONSTITUTION: A 2-color multi-color discrimination means 115 discriminates whether or not a block to be discriminated is a 2-color block on the basis of the similarity from a duplicate classification output means 114 and provides an output of 2-color/multi-color. A reference block size discrimination means 116 uses a 2-color/multi-color discrimination value and a reference block size outputted from a 2-color/multi-color criterion and a reference block size to discriminate whether or not main reference block data and a block size of a sub reference block at present is proper and provides an output of the criterion. A reference block size generating means 119 provides an output of a reference block size indicating a similar size to the block to be discriminated as an initial value. Furthermore, when a reference block size criterion outputted from the means 116 is not proper, a reference block size larger than the current size is outputted.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image dividing method and an image dividing apparatus, and more particularly to image data in which a two-color image such as a character drawing (two-color black and white) and a multicolor image such as a natural image are mixed. The present invention relates to an image dividing method and an image dividing apparatus for dividing a block into a two-color image area and a multicolor image area.

[0002]

2. Description of the Related Art In coded storage of image data in which a two-color image such as a character drawing (two-color black and white) and a multicolor image such as a natural image coexist, the image data is divided into a two-color image region and a multicolor image region. However, it is known that it is effective from the viewpoint of compression rate to perform compression using a coding method suitable for each.

Conventionally, as a method for dividing image data in which a two-color image image and a multi-color image are mixed into a two-color image region and a multi-color image region, the following documents (a) to (c) are described. The method is known.

(A) Katsuno, Endo: "Proposal of High Efficiency Hybrid Coding Scheme for Color Facsimile", Technical Report, Vol.91, No.390, IE91-96, pp.13-20 (1991) (B) Katsuno, Koike: "Adaptive Hybrid Coding Method Based on Quantization Error", Shin-Gaku Autumn Univ., D-192 (1993) (c) Katayama, Yasuda: "One encoding of text / photo mixed document Method ", Electronic Magazine, Volume 20, No. 6, pp.643-651 (1991) The method described in the above document (a) is the maximum and minimum pixel values in a block in an image divided into blocks. A method that uses a value (hereinafter, referred to as “a block determination method based on a maximum value / a minimum value”) and a method that uses a pixel value distribution within a block (hereinafter referred to as a “block determination method based on a pixel value distribution”) are combined. It is a method.

The "block discrimination method based on the maximum value / minimum value" is the maximum value (Smax) and the minimum value (Smi) in the block.
n) the difference (F; F = Smax-Smin) with a certain threshold (T)
If it is larger, it is determined that the edge exists within the block, and the block includes a two-color image.

“Block discrimination method based on pixel value distribution”
Is a method of determining that the block is a two-color image when there are one or two peak values in the distribution of the histogram of pixel values.

“Block discrimination method based on pixel value distribution”
Is used for the purpose of improving the discrimination accuracy of the “method by maximum value / minimum value”.

That is, in the "maximum / minimum value method", when there is a difference (F) between the maximum value and the minimum value in the block near the threshold value (T), it is not clear whether it is a two-color image block or not. The discrimination accuracy can be improved by discriminating using the “method based on the distribution of pixel values”.

Specifically, in the "block discrimination method based on maximum / minimum value", W1 and W2 (W1 <W2) are set instead of the threshold value (T), and the following is performed.

W2 <F then two-color area W1 <F <= W2 and one or two peak values then two-color area W1 <F <= W2 and three or more peak values then multicolor area F <= W1 Then the multicolor region The method described in the above-mentioned document (b) is a method using a quantization error generated in the quantization of transform coefficient data obtained by orthogonal transform of image data (hereinafter, "quantization of orthogonal transform"). This is referred to as "a block determination method based on an error").

This is a DCT (Descrete Cosine Tra)
The transform coefficient distribution of the orthogonal transform used for image compression such as nsform) is different between a two-color image and a multicolor image.

That is, the intensity of the low frequency coefficient becomes large in the multicolor image, and the intensity of the high frequency coefficient becomes relatively large in the two color image.

Therefore, when the same quantization table is used in the quantization of the transform coefficient for the multicolor image and the quantization of the transform coefficient for the two color image, the quantization error is different, which is particularly suitable for the multicolor image. If a different quantization table is used, the quantization error of the two-color image becomes larger than that of the multicolor image.

Therefore, when an error (E) when a certain block is orthogonally transformed and quantized using such a quantization table is larger than a threshold value (T), it is determined that the block includes a two-color image. it can.

Further, the method described in the above-mentioned document (C) is a method of extracting a black pixel by limiting black as one of the colors used for a two-color image (hereinafter, "pixel by black character detection"). And a method of discriminating a two-color image area by a filter such as Laplacian (hereinafter, referred to as “discrimination method”)
It is called a “block discrimination method using a Laplacian-like filter”. ) Is a combination method.

"Pixel discrimination method by black character detection" is 2
This is a two-color multicolor discrimination method for a document in which most of color images are composed of white and black.

When the red, green, and blue pixel values of a pixel are r, g, and b, threshold values th1 and th2 are used to determine a pixel satisfying the following expression as a two-color pixel (black pixel). .

[0018]

[Equation 1]

In the "block discrimination method by Laplacian-like calculation", the luminance value at the pixel position (i, j) is set to Y.
(I, j) = 0.299r + 0.587g + 0.144
When the threshold is b and the threshold value is th3, it is a method in which a block having pixels satisfying the following equation includes two-color pixels.

[0020]

[Equation 2]

The combination of the "pixel discrimination method by black character detection" and the "block discrimination method by Laplacian-like calculation" is performed as follows.

That is, with respect to a block which has been discriminated as a two-color image block by using the "block discrimination method by Laplacian-like calculation", two-color pixels are discriminated by using the "pixel discrimination method by black character detection".

[0023]

However, in the method described in the above-mentioned document (a), a certain block is in the two-color image area, and all pixels in the block are one of two colors. When the block is dyed, it is determined that the block has no edge at all, so that there is a problem that this block cannot be discriminated as a two-color image area.

That is, there is a problem in that a character graphic in a two-color image area larger than a block cannot be discriminated as a whole in a two-color image area.

Therefore, since the two-color image area is encoded as a multi-color image area, there is a problem that the stored information can be wasted.

The method described in the above-mentioned document (b) utilizes the property that the edges, which are often included in the two-color image, increase the level of the high frequency component of the image signal.
There is a problem that the high-frequency component in the color image area and the high-frequency component in the multi-color image area cannot be distinguished from each other, so that the accuracy of the two-color / multi-color discrimination is lowered depending on the image.

Therefore, there is a high possibility that a multicolor image area is erroneously determined as a two-color image area, and in such a case, a reproduced image is significantly deteriorated.

The method described in the above-mentioned document (C) utilizes the property that the edges of the two-color image generate the high-frequency components of the image signal, but it can be distinguished from the high-frequency components in the multicolor image region. Therefore, there is a problem that the accuracy of the two-color / multi-color discrimination decreases depending on the image.

Regarding the detection of the two-color pixels, there is a problem that only the two-color pixels of a predetermined limited color can be detected.

Therefore, there is a problem that it is necessary to determine the colors of the two-color image in advance, and the other two-color image cannot be detected.

The present invention has been made in order to solve the above-mentioned problems of the prior art, and an object of the present invention is to provide an image dividing method in which two colors are used regardless of the size of a character graphic in a two-color image area. The block in the image area is discriminated as a two-color image area, the high frequency component of the multi-color image area and the edge of the two-color image area are distinguished, and the image is divided without limiting the colors of the two-color image image. It is to provide a technology that makes it possible.

Another object of the present invention is to discriminate a block in a two-color image area as a two-color image area regardless of the size of a character graphic in the two-color image area in the image division device, and to provide a multicolor image. It is an object of the present invention to provide a technique capable of distinguishing a high frequency component of a region from an edge of a two-color image region and further dividing the image without limiting the colors of the two-color image.

The above and other objects and novel features of the present invention will become apparent from the description of this specification and the accompanying drawings.

[0034]

Of the inventions disclosed in the present application, a representative one will be briefly described below.
It is as follows.

(1) In an image dividing method of dividing image data in which a two-color image and a multi-color image are mixed into a two-color image region and a multi-color image region, (a) a two-color image region or a multi-color image region Main reference block data of the block size set by the reference block size value, which consists of aggregated data of pixel values in the block including blocks in the vicinity of the block to be discriminated to determine whether it is a color image area And (b) generating sub-reference block data having a block size set by a reference block size value, the sub-reference block data being composed of aggregated data of pixel values in a plurality of blocks not having the identified block as a center. Binarizing the reference block data and the sub-reference block data by using an average value in each block as a threshold; and (c) binarizing The step of generating a similarity value by digitizing the similarity between the overlapped portions of the main reference block data and the sub reference block data, and (d) the judged block is a two-color image using the similarity value. Determining a region or a multicolor image region to generate a two-color multicolor discriminant value, (e) the two-color multicolor discriminant value, and the block sizes of the main reference block and the sub-reference block Generating a reference block size determination value by determining whether or not the block size of the main reference block and the block size of the sub-reference block are appropriate using the reference block size value for setting; and (f). When the reference block size determination value is incorrect, the reference block size value is changed to a reference block size value larger than the current reference block size value, and And (g) when the reference block size judgment value is appropriate, the block to be judged is changed to a two-color image area or a multicolor image area based on the two-color multicolor judgment value. And a dividing step.

(2) In an image dividing device that divides image data in which a two-color image and a multi-color image are mixed into a two-color image region and a multi-color image region, a two-color image region or a multi-color image region Mainly generates the main reference block data of the block size set by the reference block size value, which is composed of the aggregated data of the pixel values in the block including the block in the vicinity of the block to be discriminated A sub-reference block generation unit that includes reference block generation means and aggregate data of image values in a plurality of blocks that do not have the discriminated block as a center, and generates sub-reference block data having a block size set by the reference block size value. And a main reference block data and a sub reference block for each block size set by the reference block size value. Binarizing means for binarizing the data using the average value in the block as a threshold value, and main reference block data and sub-blocks for each block size set by the reference block size value binarized by the binarizing means. Overlapping partial similarity output means for outputting a similarity value which is a numerical value of the similarity of overlapping parts of reference block data, and similarity for each reference block size value output from the overlapping partial similarity output means A binary multicolor discriminator that discriminates whether the block to be discriminated is a two-color image region or a multicolor image region by using a value, and outputs a two-color multicolor discriminant value; A reference block for determining whether or not the block sizes of the main reference block and the sub-reference block are proper using the two-color multicolor discrimination value and the reference block size value, and outputting the reference block size determination value. And the reference block size determination value output from the reference block size determination unit is appropriate, the block to be determined is a two-color image area or a multicolor image area based on the two-color multicolor determination value. Dividing unit for dividing into, an initial value of the reference block size value, and a reference block size larger than the current reference block size value when the reference block size determination value output from the reference block size determination unit is incorrect. And a reference block size generation means for outputting a value.

[0037]

According to each of the above-mentioned means, when the image data in which the two-color image and the multi-color image are mixed is divided into the two-color image region and the multi-color image region, the block to be discriminated is centered in the vicinity thereof. Main reference block data of a block size set by a reference block size value, and a set of pixel values in a plurality of blocks that do not have the discriminated block at the center And sub-reference block data having a block size set by the reference block size value.

Next, the main reference block data and the sub reference block data are set to 2 using the average value in each block as a threshold value.
A similarity value is generated by digitizing the similarity between overlapping portions of the binarized and binarized main reference block data and sub reference block data.

Next, using the similarity value, it is determined whether the block to be discriminated is a two-color image region or a multicolor image region, and a two-color multicolor discriminant value is generated.

Next, using the two-color multicolor discrimination value and the reference block size value for setting the block sizes of the main reference block and the sub reference block, the block size of the main reference block and the block size of the sub reference block are used. Is determined to be a reference block size determination value.

When the reference block size judgment value is incorrect, the reference block size value is changed to a reference block size value larger than the current reference block size value, and the above steps are repeated.

When the reference block size judgment value is proper, the judgment target block is set to 2 based on the two-color multicolor judgment value.
It is divided into a color image area or a multicolor image area.

In the present invention, when determining whether the block to be discriminated is a two-color image or a multi-color image, the edge of the two-color image and its surroundings are used without using the frequency distribution of the image data. Since the discrimination is made based on the similarity of the binarized image which is information, it is possible to avoid erroneously discriminating the multicolor image as a two-color image even when the multicolor image has a high frequency component. .

Further, in the present invention, since the two-color image area specialized for a special color is not extracted, the two-color image area having any color can be extracted.

Further, in the present invention, the reference block for discriminating the two-color image region or the multi-color image region is appropriately enlarged so as to discriminate, so that a wider range than the block to be discriminated can be obtained. Two-color multicolor discrimination of the used block to be discriminated can be performed, and the block in the two-color image region can be discriminated as the two-color image region regardless of the size of the character graphic in the two-color image region.

[0046]

Embodiments of the present invention will be described below in detail with reference to the drawings.

In all the drawings for explaining the embodiments, parts having the same functions are designated by the same reference numerals, and the repeated description thereof will be omitted.

FIG. 1 is a block diagram showing a schematic configuration of an image dividing apparatus which is an embodiment of the present invention.

In FIG. 1, 101 is a block dividing means, 102 is a discriminated block designating means, 103 is a division accumulating means, 104 is a main reference block generating means, 105-1.
Reference numeral 08 is a sub-reference block generation means, 109 to 113 are binarization means, 114 is an overlapping partial similarity output means, and 115 is 2
Multi-color discrimination means, 116 is a reference block size determination means, 117 is a transfer path opening / closing means, 118 is a discriminated block position generation means, and 119 is a reference block size generation means.

The block dividing means 101 divides the image data into blocks and outputs the block divided image data.

The discriminated block designating means 102 selects a discriminated block from the block divided image data outputted from the block dividing means 101 and outputs the discriminated block data.

The dividing / accumulating means 103 has a binary image buffer and a multivalued image buffer of the same size as the image data, and selects the buffer according to the two-color multicolor discrimination value while determining the block to be discriminated in the buffer. The block data to be discriminated is stored at the position indicated by the position data, and the binary image data and the multivalued image data are separately output when the discrimination processing of all the image data is completed.

The main reference block generation means 104 outputs the main reference block data of the block size indicated by the reference block size to the position indicated by the discriminated block position data.

Sub-reference block generation means 105-108
Outputs the sub-reference block data having the block size indicated by the reference block size to the position indicated by the determined block position data.

The binarizing means 109 to 113 output binarized main reference block data and sub reference block data obtained by binarizing the main reference block data and the sub reference block data by using the average value in the block as a threshold value.

The overlapping part similarity output means 114 calculates the similarity of the overlapping part of the binarized main reference block data and the sub reference block data in the area of the main reference block, and outputs the similarity value.

The two-color multicolor discriminating means 115 discriminates whether or not the block to be discriminated is a two-color block based on the similarity value outputted from the overlapping part similarity output means 114, and the discrimination result. A two-color multicolor discrimination value indicating is output.

The reference block size determining means 116 uses the two-color / multi-color determining value and the reference block size value output from the two-color / multi-color determining means 115 to determine the current main reference block data and sub-reference block block. It is determined whether or not the size is appropriate, and the reference block size determination value indicating the result is output.

The transfer path opening / closing means 117 uses the reference block size judgment value output from the reference block size judgment means 116 to divide the two-color multicolor judgment value into the divided storage means 1.
Open and close the signal path to 03.

The discriminated block position generation means 118 outputs discriminated block position data indicating the position of the discriminated block which is a block to be discriminated from now on.

The reference block size generation means 119 outputs a reference block size value having the same size as the discriminated block as an initial value, and the reference block size determination value output from the reference block size determination means 116. Is incorrect, a reference block size value indicating a block size larger than the current one is output.

FIG. 2 is a block division means 10 shown in FIG.
FIG. 3 is a diagram showing a state of a block of image data divided by 1, and each square shown in FIG. 2 is a block to be discriminated.

FIG. 3 shows the main reference block generating means 104 and the sub reference block generating means 105-1 shown in FIG.
FIG. 10 is a diagram showing a positional relationship between a main reference block, a sub reference block, and a referenced block generated by 08.

In FIG. 3, 302 is a main reference block,
Reference numerals 303 to 306 are sub-reference blocks.

Next, the principle of the image dividing method of this embodiment will be described with reference to FIGS.

In FIG. 4 to FIG. 7, one-dimensional image data will be used for simplification of description.

As shown in FIG. 5A, the two-color image has a sharp edge in the gradation characteristic with respect to the position,
It has the feature that there is a planar area around the edge.

Therefore, as shown in FIG. 5B, when the data of the main reference block (B) and the data of the sub reference block (A, C) are binarized by the binarizing means in the two-color image area. , The main reference block (B) and the sub-reference block are different even though the average value of each block is different due to the gradation characteristics with respect to the position of the two-color image in which the edge is sharp and has a constant value around it. Since the position of (A, C) is deviated, the overlapped portion between the data of the binarized main reference block (B) and the data of the binarized sub-reference block (A, C) Are almost the same data, and highly similar data can be obtained.

On the other hand, as shown in FIG. 4 (a), the multicolor image generally has no edge and has a gentle slope in the gradation characteristic with respect to the position, or even if there is an edge, the surroundings are constant. It has the characteristic of not being.

Therefore, as shown in FIG. 4B, when the data of the main reference block (B) and the data of the sub reference block (A, C) are binarized by the binarizing means in the multicolor image area. Data having low similarity can be obtained in the overlapping portion of the data of the binarized main reference block (B) and the data of the binarized sub reference block (A, C).

From this, it is possible to judge whether the block to be discriminated is a two-color image or a multicolor image based on the similarity of the binarized image.

As described above, the method of determining whether the block to be discriminated is a two-color image or a multicolor image based on the similarity of the binarized image, without using the frequency distribution of the image data, Since the edge of the color image and the information around the edge are used, it is possible to avoid misidentifying the multicolor image as a two-color image when the multicolor image has high-frequency components.

This method is also specialized in special colors.
Since the color image area is not extracted, it is possible to extract a two-color image area having any color.

However, as shown in FIG. 6A, there are cases where the structure of the image cannot be accurately grasped due to the block size of the main reference block (B), and large character graphics are present in the two-color image area. Yes, when a certain discriminated block is filled with one of two colors, the initial block size of the main reference block (B) and the sub-reference block (A, C) is the block size of the discriminated block. Therefore, in the initial stage, the block to be discriminated may be erroneously discriminated as a multicolor image block.

Therefore, as shown in FIG. 7A, the similarity in the minimum block size having no fine structure in the main reference block (B) is calculated, and when the similarity is equal to or larger than a separately determined threshold value, If the main reference block (B) is determined to be a binary image, and the degree of similarity is smaller than a separately set threshold,
Until the similarity becomes equal to or less than a separately determined threshold, as shown in (A ′), (B ′), and (C ′) of FIG. 7B, the main reference block (B) and the sub reference block (A, Expand C) and
It is determined whether the block to be discriminated is a two-color image or a multicolor image.

As a result, it is possible to discriminate the two-color multicolor of the block to be discriminated by using a wider range than the block to be discriminated, and the blocks within the two-color image area can be obtained regardless of the size of the character graphic in the two-color image area. Can be discriminated as a two-color image area.

Next, a procedure for dividing an image by the image dividing apparatus of this embodiment will be described.

(A) The image data (I) is made into block divided image data (Id) by the block dividing means 101, which is divided into squares as shown in FIG.

The discriminated block position generating means 118 is also provided.
Outputs the discrimination block position data (P) indicating the position of the discrimination block.

As a result, the discriminated block designating means 10
2 outputs the data indicated by the discriminated block position data (P) from the block divided image data (Id) as discriminated block data (Ibs).

(B) On the other hand, the reference block size generating means 119 outputs a reference block size value (Sc) having the same size as the block to be discriminated as an initial value.

As a result, the main reference block generation means 10
4 and sub-reference block generation means (105-10
In 8), the position of the discriminated block is obtained from the discriminated block position data (P), and the block size of the main reference block and the sub-reference block is obtained from the reference block size value (Sc) to obtain the image data ( I) to the main reference block (Ibm) and the sub-reference block data (Ibs1 to 1b1) having the positional relationship shown in FIG.
4) is generated and output.

(C) The main reference block (Ibm) and the sub reference block data (Ibs1 to 4) are binarized by the binarizing means (109 to 113) using the average value in the block as a threshold. Binarization means (109-113)
To the binarized main reference block data (Ibmb), and
The binarized sub-reference block data (Ibsb1-4) is output.

(D) In the overlapping part similarity output means 114, binarized main reference block data (Ibmb) and binarized sub-reference block data (Ib) in the main reference block range
The similarity (S) with sb1 to 4) is calculated and output according to the following formula.

[0085]

(Equation 3)

However, k: type of sub-reference block, bm: range of main reference block, i: image position, Ibmb (i): value at position i of binarized main reference block data, Ibsbk (i): 2 Value at position i of data of binarized sub-reference block k, Nbm: number of pixels in range bm of main reference block, ^: exclusive OR, bar x: inverse of x (x = Ibmb (i) ^ Ibsbk
(I)).

When calculating the exclusive OR,
Do it using positive logic.

That is, the exclusive OR used here is calculated as follows.

A ^ B = C A B C 0 0 0 0 0 0 other than 1 0 other than 0 1 0 other than 0 0 Inverse x (bar x) is calculated as follows.

X bar x 0 1 1 0 (e) In the two-color multicolor discrimination means 115, the similarity value (S)
Is compared with a preset threshold value (Ts) to make the following determination, and the two-color multicolor determination value (Dbm) indicating the determination result is output.

If S> = Ts then Dbm = two-color discrimination if S <Ts then Dbm = multicolor discrimination (f) In the reference block size determination means 116, the reference block size value (Sc) and the two-color multicolor discrimination value (Dbm) is used to determine whether or not the block sizes of the main reference block and the sub-reference block currently used for the determination are appropriate, and the reference block size determination value (Ds) is output.

If Dbm = two-color discrimination then Ds = appropriate if Dbm = multicolor discrimination and Sc <size of image data I then Ds = improper if Dbm = multicolor discrimination and Sc> = size of image data I then Ds = Appropriate (g) If the reference block size determination value (Ds) is incorrect, the reference block size generation means 119 outputs a reference block size value (Sc) indicating a block size larger than the present, and (B) to (f) are repeated.

(H) When the reference block size judgment value (Ds) is proper, the transfer path opening / closing means 117 is closed and the two-color multicolor judgment value (Dbm) is input to the division storage means 103.

(I) The dividing and accumulating means 103 has two buffers of the same size as the image data (I), one for binary image data and one for multivalued image data. Dbm), the discriminated block data (Ibs) is stored at the position indicated by the discriminated block position data (P) in each buffer.

When the determination is completed for all blocks, the binary image data (Ibin) and the multi-valued image data (Imu
l) and are output.

In the procedure (b), when the size of the block indicated by the reference block size value (Sc) is smaller than that of the character graphic, some of the main reference blocks in the two-color image area are Filled with one of the colors,
As a result, in (e), there is a possibility that the discriminated block, which should originally be discriminated as a two-color image, is erroneously discriminated as a multicolor image.

However, in the procedure (f),
By enlarging the reference block size for the block to be discriminated as a multicolor image, it is possible to eliminate the obstacle of misjudgment due to such a small reference block.

That is, it is possible to discriminate a block in the two-color image area as a two-color image area regardless of the size of the character graphic in the two-color image area.

The steps (b) to (g) are methods for judging whether or not the inside of the main reference block of a certain size with respect to a certain block to be judged is a two-color image. Only the information that the surrounding area becomes flat is used, and the characteristics of the frequency distribution of the image signal are not used.

Therefore, it is possible to avoid erroneously determining a multicolor image block containing many high frequency components as a two-color image.

In steps (b) to (g), the two-color image block specialized for a special color is not discriminated.

Therefore, it is possible to extract a two-color image block having any color.

The inventions made by the present inventors are as follows.
Although the present invention has been specifically described based on the above-mentioned embodiments, the present invention is not limited to the above-mentioned embodiments, and it goes without saying that various modifications can be made without departing from the scope of the invention.

[0104]

The effects obtained by the typical ones of the inventions disclosed in the present application will be briefly described as follows.

(1) According to the present invention, the reference block for discriminating the two-color image region or the multi-color image region is appropriately enlarged and discriminated. It is possible to discriminate a block in the two-color image area as a two-color image area without depending.

(2) According to the present invention, since the frequency characteristic is not used in the process of discriminating the two-color image area or the multi-color image area, the multi-color image block containing a large amount of high frequency components is processed into the two-color image. It is possible to prevent erroneous determination as a block.

(3) According to the present invention, since a 2-color image area specialized for a special color or a multi-color image area is not discriminated, a 2-color image block having any color can be extracted. Become.

[Brief description of drawings]

FIG. 1 is a block diagram showing a schematic configuration of an image division device that is an embodiment of the present invention.

FIG. 2 is a diagram showing a state of a block of image data divided by a block dividing unit 101 shown in FIG.

FIG. 3 is a diagram showing a positional relationship between a main reference block generation unit 104 and main reference blocks generated by sub reference block generation units 105 to 108 shown in FIG. 1, and sub reference blocks, and referenced blocks. Is.

FIG. 4 is a diagram for explaining the principle of the present invention, and is a diagram showing, as one-dimensional image data, a change in gradation in a multicolor image and a result of binarizing the change in each overlapping block. .

FIG. 5 is a diagram for explaining the principle of the present invention, and is a diagram showing a change in gradation in a two-color image image and the result of binarizing the change in each overlapping block as one-dimensional image data. is there.

FIG. 6 is a diagram for explaining the principle of the present invention, and is a diagram showing one-dimensional image data that an image cannot be accurately grasped depending on the block size.

FIG. 7 is a diagram for explaining the principle of the present invention, and is a diagram showing a block size determination method with one-dimensional image data.

[Explanation of symbols]

101 ... Block dividing means, 102 ... Discriminated block designating means, 103 ... Divided accumulating means, 104 ... Main reference block generating means, 105-108 ... Sub reference block generating means, 109-113 ... Binarizing means, 114 ... Duplication Partial similarity output means, 115 ... Two-color multicolor discrimination means, 302 ... Main reference block, 303-306 ... Sub reference block.

Claims (2)

[Claims] 1. An image division method for dividing image data in which a two-color image and a multi-color image coexist into a two-color image region and a multi-color image region, comprising: (a) a two-color image region or a multi-color image region. Mainly the reference block data of the block size set by the reference block size value, which consists of the aggregated data of the pixel values in the block including the blocks in the vicinity of the block to be discriminated which is to be discriminated from the image area, Generating sub-reference block data having a block size set by a reference block size value, the sub-reference block data consisting of aggregated data of pixel values in a plurality of blocks not having the identified block as a center; Binarizing the block data and the sub-reference block data by using an average value in each block as a threshold value;
(C) generating a similarity value that is a numerical representation of the similarity of overlapping portions of the binarized main reference block data and sub-reference block data; and (d) using the similarity value. Generating a two-color multi-color discrimination value by discriminating whether the block to be discriminated is a two-color image area or a multi-color image area, (e) the two-color multi-color discrimination value, the main reference block, and A reference block size determination value is generated by determining whether or not the block size of the main reference block and the block size of the sub reference block are appropriate by using the reference block size value for setting the block size of the sub reference block. And (f) changing the reference block size value to a reference block size value larger than the current reference block size value when the reference block size determination value is incorrect. And (g) repeat the steps (a) to (e), and (g) when the reference block size determination value is appropriate, the block to be determined is a two-color image region based on the two-color multicolor determination value, or And a step of dividing into multicolor image areas. 2. An image dividing device for dividing image data in which a two-color image and a multi-color image coexist into a two-color image region and a multi-color image region. Main reference that generates the main reference block data of the block size set by the reference block size value, which consists of the aggregated data of the pixel values in the block including the block in the vicinity of the block to be determined A sub-reference block generation means that includes block generation means and aggregate data of image values in a plurality of blocks that do not have the discriminated block as a center and that generates sub-reference block data having a block size set by a reference block size value. And the main reference block data and the sub reference block data for each block size set by the reference block size value. Data is binarized by using the average value in the block as a threshold value, and main reference block data and sub-data for each block size set by the reference block size value binarized by the binarizing means. Overlapping partial similarity output means for outputting a similarity value which is a numerical value of the similarity of overlapping parts of reference block data, and similarity for each reference block size value output from the overlapping partial similarity output means A binary multicolor discriminator that discriminates whether the block to be discriminated is a two-color image region or a multicolor image region by using a value, and outputs a two-color multicolor discriminant value; Block for determining whether the block sizes of the main reference block and the sub-reference block are proper using the two-color multicolor discrimination value and the reference block size value, and outputting the reference block size determination value Size determining means and the reference block size determining value output from the reference block size determining means are appropriate, the block to be determined is a two-color image area or a multicolor image area based on the two-color multicolor determining value. Dividing unit for dividing into, an initial value of the reference block size value, and a reference block size larger than the current reference block size value when the reference block size determination value output from the reference block size determination unit is incorrect. An image dividing device comprising: a reference block size generating means for outputting a value.