JPH0671308B2 - Image signal processing method - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION "Industrial field of application" The present invention relates to an image information source in which a character image and a halftone image coexist in order to appropriately perform image processing in these portions. The present invention relates to a picture signal processing method for identifying an image portion.

"Prior Art" Image processing devices that read an original using an image sensor and record or display it using a quantized image signal distinguish the respective parts when the original contains characters and photographs. Image processing is performed. This is because only the part of the character image is reproduced well by the multi-valued processing for each pixel, and the reproducibility of the halftone part such as a photograph is limited, while the pseudo halftone process such as dither processing causes a photograph etc. This is because while the halftone of 3 is well reproduced, the reproducibility of the line drawing portion such as characters is deteriorated.

As a conventional technique for distinguishing a character image portion from a halftone image portion, the method disclosed in Japanese Patent Laid-Open No. 58-3374 is representative. In this method (method), the document surface (screen) is divided into blocks composed of a plurality of pixels, and the difference in density level between the pixels having the maximum density level and the pixels having the minimum density level in each block is obtained. If the difference is larger than a predetermined value, it is determined to be a binary image area or a character image portion, and otherwise, it is determined to be a gray image area or a halftone image portion.

By the way, in such an image signal processing method, when the target pixel exists at the corner of the block, there is no data for accurately determining the image state of the target pixel, and there is a risk of making an erroneous determination.

To deal with this, in Japanese Patent Laid-Open No. 58-220563, a pixel matrix centered on a pixel of interest is set, and the average value of the pixel levels in this matrix is compared with the central pixel level. A technique for discriminating whether a pixel is a character image or a grayscale image (halftone image) is shown. According to this technique, the image state can be determined for each pixel.

[Problems to be Solved by the Invention] By the way, since a character image has a characteristic that a density gradient is larger than that of a halftone image, it is possible to discriminate an image state based on the magnitude of the density gradient. Has been done. Various methods have been conventionally proposed for obtaining the concentration gradient.
A method for obtaining the difference between the maximum value and the minimum value of the density level in a block as shown in Japanese Patent Laid-Open No. 3374 (hereinafter referred to as the first method) and the method disclosed in Japanese Patent Laid-Open No. 58-220563. As described above, a method (hereinafter referred to as a second method) of obtaining a difference between an average value of pixel levels in a block and a central pixel level is known. Further, as disclosed in Japanese Patent Laid-Open No. 50-111941, a method of obtaining the difference in density between two pixels on both sides of the target pixel (hereinafter referred to as a third method) is also conceivable.

However, each of these first to third methods has the following problems.

First, in the first method, since it is necessary to detect the maximum value and the minimum value for all the pixels in the block, there is a problem that the circuit configuration for realizing this becomes large.

Next, the second method has the following two problems.

The first problem is that it is weak against noise in the image. The reason is that the second method is, in other words, a method of obtaining the density gradient from the density level of the target pixel and the density level of the peripheral pixels. However, when the block size is small, the interval between the target pixel and the peripheral pixels is small and noise is small. This is because the fluctuation of the concentration gradient due to becomes large. Further, if the block size is increased in order to reduce the fluctuation of the density gradient due to this noise, the number of pixels to be calculated increases, and the processing becomes complicated.

Further, in the second method, the density gradient from all directions with respect to the pixel of interest is taken into consideration. However, when noise occurs on the pixel of interest, the fluctuation of the density gradient due to noise is added to the density gradient from all directions. Will be. On the other hand, it is unlikely that the density gradients from all directions will be approximately equal to the pixel of interest,
Usually, the concentration gradient is directional. Therefore, if the difference between the average value of the densities of the pixels in the block and the density of the pixel of interest is set as the density gradient as in the second method, the value becomes smaller than the original density gradient. From the above, in the concentration gradient obtained by the second method, the contribution of the original concentration gradient is reduced, while the contribution of noise is increased, and the influence of noise is emphasized.

The second problem with the second method is that the image state discrimination accuracy becomes poor, or the processing becomes complicated in order to improve the discrimination accuracy. The reason is as follows. Even in the case of a character image, the density gradient is unlikely to change abruptly on a pixel-by-pixel basis.
Discrimination accuracy becomes poor. On the other hand, if the block size is increased to increase the discrimination accuracy, the number of pixels to be calculated increases, and the processing becomes complicated.

Next, the third method has the following two problems.

The first problem is that the value of the obtained density gradient changes depending on the direction of the density gradient. This is because in the third method, if a density gradient is generated along the arrangement direction of two pixels for which a difference is to be obtained, the value of the density gradient obtained will be close to the value of the original density gradient. If the arrangement direction of the two pixels for which is calculated and the direction of the density gradient are deviated, the value of the density gradient obtained becomes smaller than the original value of the density gradient depending on the deviation. For example, when the arrangement direction of the two pixels for which the difference is obtained and the direction of the density gradient are orthogonal to each other, the obtained density gradient value is "0". Incidentally, the above-mentioned JP-A-50-111941
As shown in Japanese Patent Publication No.
Although it is conceivable to obtain the average value of the difference between the directions, even in this case, the obtained concentration gradient value changes depending on the direction of the concentration gradient. That is, if the density gradient is generated along the middle direction (oblique direction) of the two directions for obtaining the difference, the value of the obtained density gradient is close to the value of the original density gradient, but the difference is obtained 2. If a concentration gradient is generated along one of the directions, the obtained concentration gradient value will be about half of the original concentration gradient value.

The second problem with the third method is that it is weak against noise in the image. The reason is that, as described above, in the third method, the value of the obtained density gradient becomes a value smaller than the original value of the density gradient except when the direction of the density gradient is in a specific direction, and the noise This is because the difference from the concentration gradient caused by is small.

In view of such circumstances, the present invention makes it possible to accurately determine, for each pixel, a character image or a halftone image with a simple process, which is less affected by noise and is not affected by the direction of the density gradient on the image. It is an object of the present invention to provide a picture signal processing method capable of performing the above.

[Means for Solving Problems] According to the image signal processing method of the present invention, among the peripheral pixels located in the peripheral portion of the predetermined range surrounding the target pixel as the pixel to be subjected to the image processing, the positions sandwiching the target pixel. Existing in 2
The density difference between two peripheral pixels is calculated for a plurality of different peripheral pixels, and the maximum value of these peripheral pixels is compared with a predetermined threshold value. From this comparison result, the pixel of interest belongs to the character image portion or the halftone image portion. It is to determine whether or not. In this case, the two peripheral pixels for obtaining the density difference may be located symmetrically with respect to the target pixel, or may be located at the most distant positions with the target pixel sandwiched therebetween. Good.

[Examples] The present invention will be described in detail below with reference to Examples.

FIG. 1 shows a document 11 as an image information source and a window 12 made up of 3 × 3 pixel groups P _{0 to} P ₈ used for image determination. In this embodiment, the original 11 is scanned without any gap by the target pixel P ₀ to be image-processed, and the window 12 also moves along with the original. Eight pixels P _{1 to} P ₈ surrounding the target pixel P ₀
Are called peripheral pixels. The respective densities d _{1 to} d ₈ of the peripheral pixels P _{1 to} P ₈ are
It is represented as a signal level after A / D conversion of the photoelectric conversion output of the reading element. In this example, the concentrations d _{1 to} d ₈
Is represented as 6-bit (64 steps) digital data.

In this embodiment, four sets of peripheral pixels are used as the peripheral pixels that sandwich the target pixel P ₀ . These are the following combinations.

(P ₁ , P ₈ ), (P ₂ , P ₇ ), (P ₃ , P ₆ ), (P ₄ , P ₅ ) The absolute value of the density difference between the peripheral pixels of each of these groups is as follows. Become.

| d ₁ −d ₈ |, | d ₂ −d ₇ |, | d ₃ −d ₆ |, | d ₄ −d ₅ | In the present example, the largest concentration difference among these four pairs is picked up, This is compared with a predetermined threshold l _TH . If it is larger than the threshold value l _TH, it is determined that the pixel of interest P ₀ belongs to the character image portion, and otherwise it is determined to belong to the halftone image portion. This can be expressed as follows.

(I) MAX (| d ₁ −d ₈ |, | d ₂ −d ₇ |, | d ₃ −d ₆ |, | d ₄ −d ₅ |)> l _TH …… Character image (ii) MAX (| d ₁ −d ₈ |, | d ₂ −d ₇ |, | d ₃ −d ₆ |, | d ₄ −d ₅ |) ≦ l _TH …… Halftone image explain. This figure shows the signal level (density) of one line for a certain part of the document. In this figure, the signal level "1" means that the part of the document is ideal white, and the signal level "64" means that the part is ideal black. The signal level changes as a whole in the first half of scanning. In these portions, the signal level changes in multiple stages, and it can be seen that it is a halftone image area. In the halftone image area, if two adjacent points are taken, there is no great difference in their densities. Therefore smaller the density difference of the two pixels sandwiching the target pixel P _0, the pixel of interest P ₀ is likely to belong to a portion of the halftone image. If density differences are found in all directions (four directions in this embodiment) centered on the pixel of interest P ₀ and their maximum values are compared with a predetermined threshold value, the determination as to whether or not the image is a halftone image becomes more reliable. .

On the other hand, there is a sharply cut waveform in the latter half of scanning in FIG. This portion is an edge portion of a line image such as a character and is a typical character image area. In such a character image area, there is a high possibility that a large density difference will occur between two adjacent points. Therefore, if the density difference between the two pixels sandwiching the target pixel is large, the target pixel is likely to belong to the character image portion. If the density difference around the pixel of interest is centered on and examined for all groups, the determination as to whether or not it is a character image becomes more reliable.

FIG. 3 shows an image signal processing apparatus for processing an image signal using the above principle. The image sensor 21 of this device raster-scans the original document, and the A / D converter 22
Quantizes the input analog image signal 23 pixel by pixel. The digital image signal 24 output from the A / D converter 22 as 6-bit parallel data is supplied to the line memory 25 and also to the series circuit of the _first to _third latches 26 _{1 to} 26 ₃ , 3 pixels are latched. Line memory here
Reference numeral 25 is a delay memory that stores 6-bit parallel data for one line and sequentially outputs the data, and first to third latches 26 ₁ to 2
6 ₃ is a 0th and 4th to 8th latches 26 ₀ and 2 to be described later.
6 _{4-26 8} similarly to the 6-bit parallel data is a circuit that latches by one pixel a.

Latch output of the first latch 26 ₁ is supplied to the first subtracter 28 ₁ as a signal representative of the density d _1. Similarly, the latch output of the second or third latch 26 ₂ , 26 ₃ is the density d ₂ or
The second or fourth subtractor 28 ₂ , 28 ₄ is used as a signal representing d _3.
Is supplied to.

On the other hand, the digital image signal 29 delayed by one line output from the line memory 25 is the other line memory having the same configuration.
31 to the fourth, 0th and 5th latches 2
6 ₄ is supplied to the 26 _0, 26 ₅ of the series circuit. 4th or 5th
The latched outputs of the latches 26 ₄ and 26 ₅ are supplied to the _third subtractor 28 ₃ as a signal representing the density d ₄ or d ₅ .

On the other hand, the digital image signal 32 output from the line memory 31 and further delayed by one line is supplied to the series circuit of the _sixth to _eighth latches 266 to 268. The latch outputs of the _sixth to _eighth latches 26 _{6 to} 268 are supplied to the fourth, second or first subtractors 28 ₄ , 28 ₂ and 28 ₁ , respectively.

In this way, the first to fourth subtractors 28 _{1 to} 28 ₄ have the first
The target pixel P ₀ shown in the figure and four sets of peripheral pixels (P ₁ , P ₈ ), (P ₂ , P ₇ ), (P ₄ , P ₅ ), (P ₃ , P ₆ ) at the target position Will be input. At this time, the output of the 0th latch 26 ₀ is the digital image signal 33 of the pixel of interest P ₀ , which is the character processing unit 34 that performs the optimum processing for the character image and the halftone that performs the optimum processing for the halftone image. The image data is supplied to both of the processing units 35 and undergoes its own image processing.

The first to fourth subtractors 28 _{1 to} 28 ₄ calculate the difference between the signal levels of the two types of input digital image signals,
The absolute values thereof are supplied to the maximum value detection circuit 36. The maximum value detection circuit 36 detects the maximum value of these differences and supplies the value to the comparator 38 as the maximum value signal V _MAX . The comparator 38 compares this with a predetermined threshold voltage V _TH and outputs a comparison result signal 39. The comparison result signal 39 is the maximum value signal V
_{When MAX} is larger than the threshold voltage V _TH, the signal state is H (high) level, and in other cases, the signal state is L (low) level. This comparison result signal 39 is sent to the multiplexer 41.
It is used as a control signal.

To the multiplexer 41, the processed image signals 42 and 43 of the target pixel P ₀ are input in parallel from the character processing unit 34 and the halftone processing unit 35, respectively, and depending on the signal state of the comparison result signal 39, One of the processed image signals 42 or 43 is selected and output as an image signal 44. Here, the character processing unit 34 is, for example, a unit that performs ternarization processing,
In the case of this embodiment, the signal level of 64 steps is quantized into the signal level of 3 steps (white, gray, black) for each pixel of interest.
Further, the halftone processing section 35 sets threshold values individually by, for example, the dither method, and performs pseudo halftone processing. This halftone processing unit 35
Then, each dot forming the dither matrix may be binarized, or may be ternarized or multi-valued.

The image signal 44 obtained in this manner is subjected to appropriate processing according to the image state to which the pixel of interest P ₀ belongs, and good reproduction of the image can be performed in the recording unit or display unit (not shown) in the subsequent stage. become.

In the embodiment described above, the character image and the halftone image are determined by using the pixel area having the 3 × 3 matrix structure, but it is free to use the pixel area having another matrix structure. FIG. 4 shows a 4 × 4 pixel area as an example. In this case, the target pixel P ₀ is set at a position slightly displaced from the center position. When the pixel of interest P ₀ exists at the position shown in FIG. 4, the two points for measuring the density difference are not uniquely determined. For example, it may be another peripheral pixel P ₁₀ that calculates a density difference from the peripheral pixel P ₁ or P ₁₅ . When a set of peripheral pixels (P ₁ , P ₁₀ ) is selected, the distance between other pixels for measuring the density in the main scanning direction or the sub-scanning direction, for example, peripheral pixels (P ₂ , P ₁₃ ) The feature is that they are almost equal. Further, when a set of peripheral pixels (P ₁ , P ₁₅ ) is selected, there is a feature that it is easier to understand the density difference by increasing the distance between pixels.

In addition, in the embodiment, it is decided to determine whether each pixel is a character image or a halftone image in order for every pixel, but the discrimination result of a specific pixel of interest is determined as a discrimination result of a pixel group in a predetermined range including this pixel of interest. It is also effective to apply the extension as the above because of simplification of the signal processing. Further, in the embodiment, the dither processing is performed by the halftone processing unit, but it is also possible to express the halftone by the area gradation method in which the pattern of the printing ratio corresponding to the gradation is assigned.

[Effect of the Invention] As described above, according to the present invention, a plurality of sets of density differences between two peripheral pixels sandwiching a target pixel are obtained, and the image state of the target pixel is determined from the maximum value thereof. There is an effect that it is possible to determine whether it is a character image or a halftone image, and the processing is simpler than the conventional method that determines the image state from the difference between the maximum value and the minimum value of the density levels in the block.

Further, according to the present invention, when the block sizes are the same, compared to the conventional method of obtaining the density gradient from the density level of the pixel of interest in the block and the density level of the peripheral pixels,
This has the effect of increasing the interval between pixels used to obtain the density gradient and improving the accuracy of determining the image state.

Further, according to the present invention, since the interval of the pixels used for obtaining the density gradient is large as described above, the fluctuation of the density gradient due to noise is reduced and the density of the target pixel is determined from the density of the peripheral pixels excluding the target pixel. Since the image state is discriminated, the influence of noise is not emphasized when noise occurs on the pixel of interest as in the conventional method, and the effect of noise is less likely to be exerted.

Further, in the present invention, a plurality of sets of density differences between two peripheral pixels sandwiching the pixel of interest are obtained, the maximum value thereof is obtained as a density gradient, and the image state is determined from this density gradient, so the influence of the direction of the density gradient on the image is affected. There is an effect that the density gradient can be accurately obtained without being affected, and the image state can be accurately determined. Further, according to the present invention,
The value of the obtained density gradient is close to the original value of the density gradient regardless of the direction of the density gradient on the image. Therefore, the image state is determined only by the density difference between two specific peripheral pixels that sandwich the pixel of interest. The effect of being less likely to be affected by noise is obtained as compared with the case where the density difference between two peripheral pixels sandwiching the target pixel is obtained in two directions and the image state is determined from the average value thereof.

[Brief description of drawings]

FIG. 1 is an explanatory diagram showing the relationship between a pixel group used for image determination and a document, FIG. 2 is a waveform diagram illustrating the level change of an image signal for one line, and FIG. 3 is an embodiment of the present invention. FIG. 4 is a block diagram of an apparatus for realizing the image signal processing method in FIG. 4, and FIG. 4 is a configuration diagram showing another example of the configuration of a pixel group including a target pixel and peripheral pixels. 25, 31 …… Line memory, 26 …… Latch, 28 …… Subtractor, 36 …… Maximum value detection circuit, P ₀ …… Pixel of interest, P _{1 to} P ₁₅ …… Peripheral pixels.

Front Page Continuation (72) Inventor Takashi Omori 2274 Hongo, Ebina City, Kanagawa Fuji Zerox Co., Ltd. Ebina Works (56) References JP 59-163957 (JP, A) JP 58-220563 (JP, JP, 220563) A) JP-A-50-111941 (JP, A)

Claims (2)

[Claims] 1. In an image processing apparatus for performing different image processing on a character image portion and a halftone image portion other than the character image portion, a peripheral portion of a predetermined range surrounding a target pixel as a pixel to be subjected to image processing. Among the peripheral pixels located at, the density difference between the two peripheral pixels existing at positions sandwiching the pixel of interest is determined for a plurality of peripheral pixels having different densities, and the maximum value thereof is compared with a predetermined threshold value. An image signal processing method characterized by determining whether a pixel belongs to a character image portion or a halftone image portion. 2. The image signal processing method according to claim 1, wherein the two peripheral pixels for obtaining the density difference are symmetrical with respect to the target pixel.