JPS6053983B2 - Halftone signal processing device - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION When transmitting or storing halftone image information, the present invention provides processing for halftone signals that reduces the amount of information required for transmission or storage by suppressing the redundancy of pixels. It is related to the device.

In a facsimile machine, a black and white manuscript is scanned and the resulting output is binarized into an image signal sequence, which has redundancy, and it is not an efficient method to electronically transmit it as it is.

For this reason, various band compression methods have been proposed to remove redundant parts from binary image information. Although a considerable effect is observed, sufficient characteristics are not shown in the halftone photographic area. As a means for solving this problem, a band compression method using adaptive prediction uses one alcoholic acid as a reference pixel for prediction, as shown in FIG. 1.

In the figure, 60 is a predicted pixel, 61 is a reference pixel, and 62 is an arrow indicating the halftone dot interval. However, this method does not utilize the two-dimensional density gradient of a single photographic original image for prediction.

2. The selection of reference pixels is inappropriate, and redundancy is not sufficiently suppressed.

For these reasons, the transmission information compression rate is not sufficient.

The present invention improves the above-mentioned points and provides an apparatus capable of obtaining an information compression rate superior to that obtained conventionally.

The present invention is a digital band compression device characterized by the method of selecting reference pixels used when creating a predicted match/mismatch signal from a binary signal obtained by scanning a halftone dot photographic area, and the means according to the present invention. The device using this method has a higher effect of suppressing the redundancy of image information than the conventional device, and as a result, a large bandwidth and compression characteristics can be obtained.

According to the conventional method in the halftone photographic area, the compression ratio is about 5, but according to the present invention, it shows a value of 8 or more.
When the compression ratio is 8, the time required to transmit image information can be approximately 118 hours, and when recording on an image film, a memory capacity of 118 times less is required. The reasons for the high compression ratio are: (1) The method of selecting reference pixels is adopted to obtain sufficient information on where the predicted pixel is located within a halftone dot.

(2) A method of selecting reference pixels is adopted that sufficiently reflects the density gradient information of the original image in pixel prediction. (3
) Furthermore, within a reference pixel block consisting of a plurality of reference pixels, moving a reference pixel located in the shadow of one reference pixel from the predicted pixel indicating the position of the same block or the center to a position where it is not in the shadow. .

Due to reasons such as.

The present invention will be explained below using the drawings.

FIG. 2 is a block diagram showing elements necessary for implementing the present invention in a facsimile transmitter. The image information of the calligraphy original 1 is illuminated by a scanning light spot from a light source 2, and the reflected light from this is sent through a lens 3 to a photoelectric conversion element 4, where it is converted into an electrical signal corresponding to the shading of the calligraphy or drawing. This electrical signal is binarized into O or 1 for each pixel by a binarizer 5 and stored in a memory 6 in a form corresponding to the pixel arrangement. A reference pixel for creating a prediction signal of a prediction pixel is selected from among the image signals stored in this memory. From now on, for the sake of simplicity, we will associate white with 0 and black with 1, and call white or black a color. The reference pixel is usually selected from the vicinity of the predicted pixel, but depending on the properties of the image to be predicted, it may be selected from a distant location, as in the case of prediction of halftone photograph pixels described in the prior art. It's okay.
In general, the higher the number of reference pixels, the better the prediction matching rate, but due to hardware limitations in the current technology, 4 to 1 is usually selected. When a reference pixel is determined as a relative position with respect to the predicted pixel, the predicted pixel is predicted by the predictor 7. The predictor 7 has a table to refer to, and the predicted pixel is determined by P% from the black and white combination of reference pixels.
Determine whether it is white or black with probability (P>50%). This result is converted into an O or 1 signal and sent to the comparator 8. The binarized image signal is stored in the memory 6, while the predicted image signal is also sent to the comparator 8.

The comparator 8 compares the received predicted value with the actual value, generates a coincidence signal, for example O, when the predicted value matches the actual value of the image, and generates a mismatch signal, for example 1, when the predicted value does not match. Send to 9. The match/mismatch signal sequence generated by the comparator 8 is encoded by the encoder 9. Various encoding methods are used for this. For example, there is a method of encoding using a known Huffman code. The encoded image signal is temporarily stored in a buffer memory and sent through a modem as necessary.
sent to the receiver. At the receiver, the image signal is reproduced through the reverse process of the transmitter, and then sent to the printing device.
The image is played. Here, the method of selecting a reference pixel from the viewpoint of a predicted pixel, which is the main point of the present invention, will be explained in detail.

Figure 3 shows halftone dots 22 and 2 of a commonly used 45 halftone photograph (the concept is the same for 90 and other halftone dot arrangements).
A coordinate axis 24 whose origin is the predicted pixel 21 indicated by marks 3 and 9.
, 25, 26, 27, 28, 29, 30, and 31 at a certain time. Coordinate axes 24, 25, and 24 are indicated by solid lines, with the direction of arrow 32 as the main scanning direction.
The interval Em between and 26 is an integer N. of the pixel width Wm in the main scanning direction. ．．
times the halftone dot spacing L. ．． The value is extremely close to . For example, if the pixel width in the main scanning direction is W. ．． = 27.6pm, then 6 per inch? , for 45W halftone dot, Nm=20
(even number), that is, E. ．． =552.0pm,. Wm=
In the case of 29.1 μm, Nn. = 19 (odd number), i.e. 1
．． ．． =552.9μm, and L. ．． ,=552.6μ
It is m. The distance between the coordinate axes 27, 28 and the coordinate axis 24 indicated by broken lines is 112 en. It is. Similarly, with the direction of the arrow 33 as the sub-scanning direction, the interval '8 between the coordinate axes 29 and 30 shown by solid lines is an integer Ns times the pixel width Ws in the sub-scanning direction, and is also a value extremely close to the halftone dot interval L6. It is.

For example, if the pixel width in the sub-scanning direction is Ws = 55.3 μm, Ns = 10 (even number), or '5 = 553.0 pm.
、． When Ws=61.4 μm, Ns=9 (odd number), that is, 1s=552.6 μm. Coordinate axis 3 indicated by broken line
1 and the solid coordinate axes 29 and 30 are 112 eS. In the above case, the pixel widths in the main scanning direction and the sub-scanning direction are changed, but even if they are the same, this does not affect the main idea of the present invention. Usually N. If n and ns are taken as even numbers, the intersection of the coordinate axes indicated by the broken line in Figure 3 allows pixels that can be taken in the same way as the predicted pixels to be placed relative to the halftone dot, but Nm, n
When s is an odd number, as described above, a pixel translated by 1h pixels is taken. When Nm and ns are even numbers, the intersection points of the coordinate axes 34, 35, 36, which are called center points,
The pixels located on 37, 38, 39, and 40 (shown) are called center pixels, and for simplicity, the same number is given to the center point and the center pixel in this specification.

Also N. ．． , n8 is an odd number, one pixel among the pixels touching the center point, that is, the intersection points 34, 35, 36, 37, 38, 39, and 40 of the coordinate axes is called the center pixel. As shown in FIG. 3, the center pixel 34 is located on the intersection of the coordinate axes 26 and 29. Similarly, the center pixel 35 is located on the coordinate axis 25.
and 29, the center pixel 37 is located on the intersection of 25, 30, and 38
are 26 and 30, 39 is 28 and 31, 40 is 27 and 3
There is a relationship of 1. The distance and positional relationship between the center pixel and the origin pixel (predicted pixel) are always constant. The origin pixel, that is, the predicted pixel moves pixel by pixel in the scanning direction as time passes, and the center pixel also moves accordingly, changing its position relative to the halftone dot. The reference pixels are arranged around the origin pixel or the center pixel and connected thereto to form an island-like pixel block. This will be called a pixel block, and the block number will be called the number of the center pixel. From FIG. 3, the pixel block 21 includes the origin pixel. Further, the number of pixels included in one pixel block is called block size BS. The size of each block is ±' in the main scanning and sub-scanning directions from the center point of each block. /4, a range with a width of ±F9/4. Fourth
The reference pixel array of the present invention will be explained in detail with reference to the drawings. The higher the number of reference pixels, the better the prediction accuracy, but in actual devices, it is not possible to increase the number infinitely, and 4 to 1 exposure level is usually used. When the number of reference pixels is small, such as 4 to 8, it is sufficient to reduce the selection of reference pixels corresponding to the center point that is the intersection of each coordinate axis. When selecting the center point, select the center point 39 when there is one point, the center point 39, 40 when there are two points, and the center point 39 when there are three points.
40, 35, 4 points, center point 39, 40, 35, 3
It is recommended to select as shown in 6. Here, the case of one alcoholic acid will be mainly explained. In FIG. 4, 21 is a predicted pixel, 2
1A~21D134C135,35E136,37,3
8, 39, and 40 are reference pixels, and 22A, 22B, and 22C are halftone dots.

(1) Around the predicted pixel 21, the reference pixel 21A,
Take 21B, 21C, and 21D.

On the other hand, reference pixels are not set to the left of 21A, the upper left of 21B, the upper left of 21C, the upper right of 21D, etc. This has not been taken into consideration in the prior art, and when the total number of reference pixels is limited, it is not a good method to arrange reference pixels at such locations. The reason for this is that experiments related to the present invention have revealed that pixels in such locations have a small amount of information for prediction. That is, in a block including a predicted pixel, a pixel in the shadow of one reference pixel has a small amount of information, so it is better to place the reference pixel somewhere else than to place it here. The role of the reference pixel group of the block including the predicted pixel 21 is to know the surrounding situation of the predicted pixel.

In other words, how many halftone dots have one predicted pixel? Its main role is to know if it is in the correct position. If all the reference pixels are black, the predicted pixel is estimated to be inside the halftone dot, and if all the reference pixels are white, the predicted pixel is estimated to be outside the halftone dot. When part of the reference pixel is black and part is white, predicted pixel 1 is located around or near the halftone dot. (2)
The reference pixels 35, 36, 37, 38, 39, and 40 are located on the coordinate axes intersection points and are located at the same position as the predicted pixel across the halftone dot interval, and are used to also know the density gradient of the original image. be. In FIG. 4, when the original image becomes darker from the upper left to the lower center, the size of the halftone dot increases from 22A to 22B and 22C. When the reference pixels are located around the halftone dot, these reference pixels will have density gradient information. For example, reference pixel 37 is white and reference pixel 39
If the predicted pixel 21 is black, it can be said that there is a high probability that the predicted pixel 21 is black. (3) The reference pixels 35 and 35E are as described in (1) and (2) above.
) can be said to have the characteristics of both.

(4) The reference pixels 34C and 35E often play a role in detecting halftone dot location information.

Although the roles have been divided as described above, each reference pixel often has one role or the other, and in many cases, it often has both roles.

Here, the main points of the present invention will be summarized from another angle.

(1) Based on the predicted pixel, the center pixel of the reference pixel block is set at the halftone dot interval or 112 in both the main scanning direction and the sub-scanning direction, and these are the actual halftone dot arrangement as shown in Figure 3. Match the relationship.

In other words, consider a coordinate system with the center of the predicted pixel as the origin and the main scanning direction and sub-scanning direction as the A center point is set at a plurality of mutually different points given by integers (Mk, N, which are even or odd numbers at the same time in one coordinate), and a center pixel is set at a position corresponding to this center point.
(2) If there are four reference pixels for a block including a predicted pixel, the reference pixels are the left, upper left, upper, and upper right of the predicted pixel.

If there are three, exclude the top left.

There is no problem with adding four or more, but it is not as effective as adding more.

(3) In FIG. 4, up to five reference pixels can be taken for the blocks whose center pixels are 35, 39, and 40.

There is no problem with using more than this, but the effect is small. Reference pixel is 5
The pixel arrangement within a block in the case of 1 is shown in FIGS. 5A and 5B.
Centering on the center pixel 45, 46A, 46B on the top, bottom, left and right
, 46C, 46D are diagonally connected to 45, 46E,
46F, 46G. There is one with 46H on it. Figure 5A
, b can be used for multiple blocks, but it is more effective to use different arrangements.

For example, FIG. 5a is used for block 39, and FIG. 5b is used for block 40. When the number of reference pixels in one block is 4 or less, they are removed from those shown in FIGS. 5A and 5B as necessary.

For example, when there are 3 pixels, 46A: 45: 46C, 46B:
45:46D. As in the case above, these function together when used in blocks 39, 40, etc.
direction) so that they complement each other. The same idea applies when the number of reference pixels in one block is 2 or less.

(4) In a pixel block that is far from a block including the center pixel, or in a pixel block that is in the shadow of a certain pixel block when viewed from the center pixel, only one pixel of the center pixel is placed as a reference pixel.

FIG. 5c shows pixels 47A, 47B, 47C, and 47D at grid points 39 and 40 when both Nm and ns are odd numbers.

When there are 4 pieces, as shown in Figure 5c, when there are 3 pieces, a suitable one of them is defected, when there are 2 pieces, a combination of 47A, 47C or 47B, 47D is made, and when there is 1 piece, a suitable one is made defective. Take pieces.

N. ．． , ns, when one of them is an even number and the other is an odd number, the reference pixels 48A and 48B are arranged symmetrically about the center pixel for the even number and line symmetrically for the odd number, for example, as shown in FIG. 5d. . Specific examples of pixel arrangement are shown in FIGS. 6A to 6D to help understand the above-mentioned points, but since the scope of the present invention is what has been described above, it is limited to the illustrated figures only. It is not something that will be done.

The color of the predicted pixel 21 is given as a color with a statistical probability of 0.5 or more from the combination of the colors of the reference pixels determined as described above.

This statistical probability is obtained through actual measurement using an image similar to that used for prediction, and is prepared in the form of a logical formula or numerical table in the predictor 7 shown in Fig. 2, and every time one pixel is measured. The predicted value is calculated from the combination of predicted pixel colors. As described above, the present invention uses the predicted pixel in the digital band compression device as a reference, and calculates the pixels at the halftone dot interval or at the IP position of this (if the number of pixels between the halftone dots is an odd number) in the main scanning and sub-scanning directions. This is a halftone signal processing device that uses the center pixel of the reference pixel block (such a position) as the center pixel of the reference pixel block, and places reference pixels around this and the predicted pixel, and predicts the two-dimensional density gradient of the image. It can be used for many purposes, and redundancy can be sufficiently suppressed to obtain an excellent information compression rate.

[Brief explanation of the drawing]

FIG. 1 is a diagram showing pixels used in conventional prediction, FIG. 2 is a block diagram showing the basic configuration of a halftone signal processing device in an embodiment of the present invention, and FIG. 3 is a diagram showing pixels and halftone dots. FIG. 4 is a diagram showing the relative positional relationship, FIG. 4 is a diagram explaining the functions of pixels used for prediction, FIGS. 5 a to d are diagrams showing examples of arrangement of pixels used for prediction, and FIGS. FIG. 3 is a diagram illustrating an example arrangement of pixels used for prediction. 4...Photoelectric conversion element, 5...Binarizer, 6...E memory, 7...Predictor, 8...
... Comparator, 9 ... Encoder.

Claims (1)

[Claims] 1. When predicting pixels in a halftone image area using an adaptive prediction method, the pixel width in the main scanning direction is W_m, the pixel width in the sub-scanning direction is W_s, and these are multiplied by integers n_m and n_s, respectively. The values are l_m(=W_m×n_m), l_s(=W_
s×n_s), l_m and l_m are values close to the halftone dot spacing L_m and L_s in the main scanning and sub-scanning directions, and the center of the predicted pixel is the origin, and the x-axis is the main scanning direction and the sub-scanning direction, respectively. , a means for taking a first reference pixel around the predicted pixel on the origin, considering a coordinate axis to be the y-axis;
Coordinate points where _k, N_k are 0 or positive integers (±1/
2l_m・M_k, ±1/2l_s・N_k), and select the pixel on the selected point, this pixel, the pixels around this pixel, or both pixels as the second a halftone dot signal processing apparatus, comprising means for using the first and second reference pixels as reference pixels, and predicting the prediction pixel using the first and second reference pixels. 2. The halftone signal processing device according to claim 1, wherein pixels to the left, upper left, upper, and upper right of the predicted pixel are selected as the first reference pixels. 3. Within each reference pixel block formed as a collection of reference pixels around each center point with the origin and the selected coordinate point as the center point, a reference pixel block is placed in the shadow of one reference pixel when viewed from the center point. The halftone signal processing device according to claim 1, characterized in that no pixels are taken. 4. The halftone signal processing device according to claim 1, wherein the number of reference pixels is increased for a reference pixel block close to the predicted pixel, and the number of reference pixels is decreased for a reference pixel block far from the predicted pixel. 5 If the selected coordinate point is one point, the coordinate (-1/2l
2. The halftone signal processing device according to claim 1, wherein the halftone dot signal processing device selects a point of _m, +1/2l_s). 6 If there are two coordinate points to be selected, the coordinates (-1/2l
__m, +1/2l_s) and (+1/2l_m, +1
2. The halftone dot signal processing device according to claim 1, wherein two points of /2l_s) are selected. 7 If there are 3 coordinate points to be selected, the coordinates (-1/2l
__m, +1/2l_s), (+1/2l_m, +1/2
2. The halftone signal processing device according to claim 1, wherein three points, l_s) and (-l_m, 0), are selected. 8 If there are 4 coordinate points to be selected, the coordinates (-1/2l
__m, +1/2l_s), (+1/2l_m, +1/2
4 of l_s), (l_m, 0) and (0, +l_s)
2. The halftone signal processing device according to claim 1, wherein a dot is selected. 9. The halftone image processing device according to claim 1, wherein one pixel corresponding to one coordinate point is taken as a reference pixel.