JPH02268073A - Picture reducing device - Google Patents

Abstract

PURPOSE:To save a reduced picture even at the mixture of a line drawing dither or the like by applying filtering and subsampling to a binary picture to reduce the signal, referencing reduced picture element and elements around the reduced picture element and correcting the result of reduction to a specific pattern. CONSTITUTION:A read original picture signal in a facsimile equipment is binarized by a comparator 230 via a filter section 220, subsampled and reduced without loss of thin lines or the like. The reduced picture element and elements around the reduced picture elements are referenced by an exemption processing section 260, the reduced binary picture element is corrected based on the statistical adaptive processing to a specific pattern and outputted via a selector 240 and the reduced picture is saved even during the mixture of line drawing dither or the like.

Description

[Detailed Description of the Invention] [Industrial Application Field] The present invention applies to an image reduction device for binary still images, etc. In a facsimile machine, one image is sequentially scanned in the raster direction and encoded and transmitted. A method is being adopted. With this method, it is necessary to transmit encoded data for all images in order to grasp the entire image, so the transmission time is long and it is difficult to adapt to image communication services such as image database services and Videotex. Ta.

Therefore, hierarchical encoding has been considered in order to quickly grasp the entire image. FIG. 14(a) shows an example of conventional hierarchical encoding.

101 to 104 are each 1. Repair, I! 105 to 107 are respectively a frame memory for storing a reduced image of 1/8, a reduction unit that generates a reduced image of 1/8,
Encoders 108 to 111 encode reduced images of 178°h, tsuba, and l, respectively.

The reduction unit 105 reduces the image from the frame memory lot in both the main scanning and sub-scanning directions using techniques such as subsampling, and generates an image of the station size.
Store in 02. Furthermore, the moon size image is reduced by the reduction unit 106 to create a % size image and stored in the frame memory 10.
Similarly, a low-resolution image of l/a size is created by the reduction unit 107 and stored in the frame memory 104.

Encoding is performed by transmitting codes sequentially starting from low-resolution images, so that a rough overview of the entire image can be quickly grasped. In the example of FIG. 14(a), the image is reduced to 178 pixels in both the main scanning and sub-scanning directions, and the encoding is I/8,
弼、IC! This is an example in which the images are processed in the order of the X size and transmitted in this order. Frame memory 104 is used to encode 1/8 image.
The encoder 1 sequentially scans the 178 images stored in the
For electronic images, entropy coding such as arithmetic coding is performed with reference to the pixel of interest to be encoded by 01 and surrounding pixels, the surrounding pixels of the pixel of interest from the frame memory 104 and 1/8 from the frame memory 104 Encoding is performed by the encoder 109 by referring to surrounding pixels of the image, thereby increasing the encoding efficiency. Similarly, for the electric image in the frame memory 102, the electric image in the frame memory 103 is referred to, and for the original size image in the frame memory 101, the encoder 110, 1 refers to the electric image in the frame memory 102.
11, encoding is performed respectively.

Further, reduction of binary images is also performed in devices other than still image communication devices. For example, this is the case when images are output from the same image database to printers with different output resolutions.

A binary image read at 400dpi is read at 300dpi
Alternatively, when outputting to a 200 dpi printer, it is necessary to reduce the image to 3/4, )S in both the vertical and horizontal directions.

Conventionally, when such reduction is performed, sub-sampling is used to thin out pixels at regular intervals, or.

A method is used in which the signal is subjected to a low-pass filter, then binarized and subsampled.

[Problems to be Solved by the Invention] In the hierarchical encoding method, early transmission of the entire image is possible by sequentially transmitting reduced images in order of lower resolution as described above. Therefore, it is necessary to leave information on the reduced low-resolution image so that the entire image can be easily understood.

When such reduction is performed using conventional methods, there is a drawback that information is lost. FIG. 14(b) is an example of subsampling the x-marked pixels of the original image (1) to obtain a reduced image (2) in the vertical and horizontal directions.

In the case of only subsampling, if one line L is located in the middle of the sampling points (marked with an x in the figure) as shown in the figure,
This line disappears upon shrinkage. In order to eliminate such drawbacks, a method of performing subsampling after filtering has been considered. An example of this is shown in Figure 14(c).
The x marks in FIG. 14(C) are sampling points. In the example of FIG. 14(C), before performing subsampling, a 3×3 low-pass filter having coefficients as shown in (3) of 5g14(C) is applied, and the filter output is binarized. For example, if the filter output is 8 or more, it can be binarized and defined as O if it is less than 1.8. However, even in the method using filtering, the drawback that the line disappears when the single vertical line L2 in the original image is between subsamplings in the example of FIG. 14(C) cannot be improved.

Therefore, in a system that repeats reduction many times, it is necessary to save a line with a width of 1 pixel, and in a low-resolution image, a line with a width of 1 pixel must be saved.
Eventually the line will disappear. Therefore, it is necessary to preserve a thin line such as a one-pixel line regardless of the sampling point.

[Means and effects for solving the problems] The present invention has been made in view of the above drawbacks. In other words, in the conventional method for reducing binary images, problems such as disappearance of thin lines in line drawings and loss of density and information in dithered images have arisen.The present invention provides filtering of binary images and exception processing to compensate for this problem. This method solves the above-mentioned problem by storing a reduced image.

Furthermore, the image characteristics are determined through statistical processing and the image is reduced adaptively, and the above-mentioned problems can be solved even for images containing line drawing dither etc.

[Embodiment] FIG. 1 is an example of a binary image hierarchical encoding device to which the present invention is applied. 1 is a frame memory FMI that stores one screen of the original image, 2 is an image reduction unit RD for generating a reduced image, and 3 is a generated reduced image (in this embodiment, vertical, horizontal, %, I/ Frame memory FM2.4 stores the reduced image (image reduced to 1/8
This is an encoding unit that encodes and transmits a reduced image using an arithmetic code method.

First, an image signal is stored in the frame memory 1 via an image input device (not shown) line 10. An image signal is output from frame memory 1 to line 11 for each rask.

This signal is selected by the selector 5 and inputted to the image reduction section 2 from the line 51. Here, the image is reduced vertically and horizontally by %.

Does selector 5 reduce the original image? Frame memory 3
Select whether to further reduce the reduced image stored in . When a reduced image (repair or service image) is selected by the selector 5, the reduced image is output from the frame memory 3 to line 31, a signal is selected by the selector 5, and input to the reduction unit 2 via line 51. .

One image reduced vertically and horizontally by the reduction unit 2 is stored in the frame memory 3 from the line 21. After storing the image obtained by reducing the original image into the frame memory 3 through the above process, the H reduced image is outputted from the frame memory 3 to the line 31 for each rask, and then the H reduced image is outputted to the line 31 by the selector 5 through the line 51 and then reduced by the same process. An image is formed in the frame memory 3. Similarly, 1j8 images are stored in the frame memory 3.

In this way, the stored reduced images are sequentially encoded at 1/89% by the encoding unit starting from the 1j8 image, and the 0 image signals, which are hierarchically encoded in the order of the original image, are encoded on lines 15, 14, and 14, respectively.
13.12, it is input to the code section 4. The encoding unit 4 uses an arithmetic code method, and encoders 108 to 108 in FIG.
The encoding process is performed in a similar order to 110 and the encoded output is output via line 17.

Details of the reduction unit 2 are shown in FIG. 2, 210, 211, 21.
Reference numeral 2 denotes line memories each storing pixels for one line, and input to the filter section 220 from lines 21:1, 214, and 215, respectively. Details of the filter section 220 are shown below.

Figure 3 shows the coefficients of a 3x3 filter used in the filter section 220. In this embodiment, a low-pass filter is used, and the weighting coefficient of the center pixel is set to C (C is 4 as standard). Various weighting factors are given. If the density value of the center pixel is D1j (i=I, jml~N, MN is the image size in the horizontal and vertical directions), the output density W of the filter is W= (Di-f, j-1+2Di, j -1+
[)i+l,j-1÷2Di-1,j+CD i,j2
Di÷l, j+Di-1, j+1+2Di, j+l+D
i+l.

j-1) (1) The output W221 of the filter section 220 is binarized by the comparator 230 using a threshold value T (T=8 as standard in this embodiment). It is binarized as in this case. Binarized output is selector 2
40 and reduced to vertical and horizontal co-stations by the sub-sampling unit 250.

FIG. 4 is an explanatory diagram of subsampling.

By extracting every other pixel data indicated by diagonal lines in the figure in the main scanning and sub-scanning directions, a sub-sampled image of the local size (% in area) is formed. This can be easily realized by latch timing of pixel data. The output of the sub-sampling section 250 is sent to the line memory 2.
The signal is outputted via 70 to the frame memory 3 shown in the first diagram.

By the way, the low-pass filter (coefficients in Figure 3) is 1 or more.
Although the surrounding pixel density is preserved and local pixel reduction is performed by subsampling, lines such as the first pixel may disappear due to the phase of the subsampling position. An example of this is shown in Figure 5. In Figure 5, c=
4. After applying a low pass filter with T=8, 3x3
This is the result when subsampling the center of the pixel.

(a) shows a case where a single black vertical line does not pass through the center of 3×3 pixels, and the subsampling result is white and one line disappears.

Similarly, the diagonal black line in (b), the central white line in (C), etc. disappear. Therefore, it is necessary to store information such as lines regardless of the sampling position.

Therefore, as shown in the second figure, an exception processing section 260 is provided to perform exception processing in addition to reduction processing by filtering and subsampling, thereby preserving information such as thin line edges and isolated points. FIG. 6 shows the configuration of the exception handling section 260.

The exception processing section 260 includes a statistics processing section 600 and an exception pattern processing section 610. Each signal line number is the same line as the same number in FIG. The statistical processing section 600 examines the black and white distribution of pixels around the pixel of interest to be reduced, and outputs an exception pattern selection signal 601 to the exception pattern processing section 61G to determine one exception pattern. 7th
An example of this is shown in the figure, where the main scanning direction (horizontal direction) is i, the sub-scanning direction (vertical direction) is j, and the pixel of interest to be reduced is x(i, j).

x (11J) is the median value of pixels filtered by a 3×3 low-pass filter (Equation (1) Di+j)
In this embodiment, the statistics (distribution) of the surrounding 5×5 pixels are calculated. The characteristics of the image, such as characters, line drawings, dithered images, positive images, negative images, etc., are determined based on the statistics, and an appropriate exception pattern is selected. Determination as to whether the image is a positive image or a negative image is performed as follows.

If the sum of the densities of the 5×5 pixel area is S, then j==-21
==-2 @ (x (i, j)=Oorl Q
:white pixel l:black pixel) If STN (=18), there are many black pixels, so it is determined that the image is a negative image.

Furthermore, a dithered image or other halftone-expressed image is determined as follows. That is, a 5×5 pixel array is divided into five stripes of 5 pixels each horizontally, and black and white inversions in each stripe are counted. Assuming that the sum of each count value is CN, if CN>16, the number of 1 inversions is large, so it is determined that the image is a 5-dither image or a halftone image. This number of black and white inversions can be easily configured using a counter.

In the exception pattern processing unit 610, line memories 21O to 2
Referring to the 3x3 pixels in the reduced image from 12, the subsampling 250, and the 3 pixels already reduced from the line memory 270, if the result by the filter is not favorable, outputs the 1 exception pattern signal 21i1 and the exception pattern selection signal 262. However, the selector 240 can output an exception pattern signal instead of the filter output.

FIG. 8 shows reference pixels in the exception pattern processing section 61G. (1) represents the image after reduction processing, (2) represents the image before reduction processing, and e represents the target pixel of the image to be reduced, a, b, c, d, f.

g*h*i are surrounding reference pixels; these are 0 pixels abc, which are the same as the pixels applied to the filter section 220;
, def, and ghi are the line memories 210 and 2, respectively.
11, 212 through lines 213, 214, 215. On the other hand, X is the pixel resulting from the reduction, and A,
B and C are already reduced pixels, lines 251 and 271
input through.

FIG. 9 shows an example of selecting one exceptional pattern when the characteristics of an image are determined by the statistical processing unit 600. 9th
Patterns (1), (2), and (3) in Figure (a) are cases in which the output is different for a positive image and a negative image. In such a case, if the image is a positive image of four characters or line drawings, it is preferable that the reduced image be black (4).

Conversely, in the case of a negative image, white (5) is a preferable example. Therefore, for the pattern shown in FIG. 9(a), the exception pattern processing section 610 outputs black (1) for the exception pattern signal 261 in the case of a positive image and white (0) in the case of a negative image.

At this time, a signal indicating an exception is output as an exception pattern selection signal 262, and the selector 240 in FIG. 2 selects the exception pattern 261 from the exception processing unit 260, resulting in a reduced pixel result (pixel ).

Pattern (I) in Figure 9(b) is an example of whether black (2) is better for dithered images in terms of density preservation, or white (3) for non-dithered images such as line drawings. It is. (a
) and (b), the reduced pixel C (FIG. 8) may be white or black.

As described above, in the reduction method of this embodiment, the reduced image obtained by subsampling after filter processing is corrected using the exception processing pattern. Additionally, exceptional patterns are selected through statistical processing and adaptive reduction is performed according to the image quality.

<Second Embodiment> A low-pass filter as shown in FIG. 3 was used in the filter section in the embodiment of the reduction method of the present invention. This is a case where the same coefficient as that of the low-pass filter is applied to the vicinity of the pixel of interest to be reduced by applying a feedback coefficient of 3. In this case, if W is the output density of the filter, it can be binarized as black when W>5 and white when W<5. A similar effect can be obtained by correcting this result for exception processing.

<Third Embodiment> In the above embodiments, the statistical processing section adaptively selected the exception processing pattern depending on the nature of the determined image, but the filter coefficients may also be selected based on the results of the statistical processing section. can. An example is shown in Figure $11, 171
1 is a first filter, 172 is a second filter, 173 is a selector, and 174 is a statistical processing unit. First filter 171
A low-pass filter (Fig. 3) is used for the second filter 17.
2, a recursive filter (Fig. 10) is used. If the judgment result by the statistical processing unit 174 is a dithered image, the second filter 172 is selected, and if the judgment result is a line drawing, the first filter 171 is selected, thereby improving the image quality of the filter output. In addition, filter exception processing can be adaptively selected based on the results of the statistical processing section.

<Fourth Example> Incidentally, the judgment of a negative image may be performed in the statistical processing section as follows. FIG. 12(a) is an image to be reduced, and a, b, c...i are This is the same pixel as in FIG. 8(b). The pixel to the left of abc...i is a pixel that has already been referred to for reduction. Figure (b) shows the pixels that have been reduced.
X is the pixel that is currently being reduced. Further, ABC is a pixel that has already been reduced and is the same pixel as ABC in FIG. 8(1).

In FIG. 12(a), there are three lines (L; Ll) in the main scanning direction.

12. Count the black runs or white runs of the images that have already been referenced in >13. In other words, the white run of each line up to now is W (411) W (also 2) W (see 3) black run (also 1)
B(12)B i3) then B(11) aandB(412) a andB13
) If α − (4-1), it is determined to be a negative image, and W(!Lt) βandW(ffi2) βandW(
Statement 3) If nβ·−(4-2), the respective processes can be performed by determining that the image is a positive image.

Normally, the frequency of use of negative images is lower than that of positive images, so α>>β is effective. Such statistical processing is performed dynamically. In addition, the initial value is a positive image, (4
Positive image processing is performed until -1) is satisfied, and when (4-2) is satisfied, zero or more negative image processing is dynamically performed from that location.

Furthermore, the same operation can be performed using the reduced image shown in FIG. 12(b). In other words, the two lines Ll and L of the reduction part
2, and the white run black run W(Ll)W(L2
) and B (Ll)B (L2), similarly (4-1)
It can be determined as in (4-2). Furthermore, when a negative image is determined, the negative image can be processed in the same way as the positive image by using pattern inversion before and after processing the positive pattern.

FIG. 13 shows an example of the configuration of an exception handling section using this process. Each signal line in FIG. 13 is the same as that in FIGS. 2 and 6. 600 is a statistical processing section, as mentioned above,
The attributes of the image are determined based on the values of the black run and white run. 6
20, 630 is an image inversion unit, and the image inversion unit 620 is the 12th
abcd'efghi in diagram (a) and ABC in (b)
When the statistical processing unit 600 determines that the input image is a negative image, the black and white image is inverted.

Further, the image inversion unit 630 converts the black and white of the reduced pixels (X in FIG. 18(b)) processed by the exception pattern unit to the statistical processing unit 630.
Similarly, in the case of a negative image, it is inverted according to the signal 601 from 0. The operations of other parts are exactly the same as those shown in FIG.

That is, even if the statistical processing unit 600 determines that the image is a negative image, it can be reduced using the same pattern as a positive image, resulting in improved image quality when reducing a negative image.

[Effects of the Invention] As described above, according to the present invention, information such as thin lines or dithered images that disappear during reduction in conventional binary image reduction can be preserved even if the information is present in the reduced image. Further, even if reduction is repeated, line image information is preserved because one pixel line is preserved. Furthermore, since the size is adaptively reduced through statistical processing, it is possible to save each image information even in a mixed image such as a line drawing dither.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an outline of an encoding device. FIG. 2 is a block diagram of the reduction section. Figure 3 is a diagram showing the coefficients of the low-pass filter, Figure 4 is a diagram showing the subsampling method, Figure 5 is a diagram showing an example of thin lines disappearing due to the low-pass filter, and Figure 6 is a block diagram of the exception handling section. . FIG. 7 is a diagram showing a statistical processing method, and FIG. 8 is a diagram showing exception processing reference pixels. FIG. 9 is a diagram showing an example of exception handling selection by statistical processing. 10 is a diagram showing an example of a recursive filter, FIG. 11 is a diagram showing an example of selecting a filter, FIG. 12 is a diagram showing a pixel arrangement, FIG. 13 is a block diagram of another exception handling section, Figure 14 shows the conventional configuration, where 1.3 is a frame memory, 2 is a reduction section, 4 is an encoding section,
5 is a selector, 220 is a filter section, and 260 is an exception handling section. Mawariwaki Ward □Wal Visit C Figure 5 Total 2 Tsuyoshi) (b) Sae 1B! Rui (a)

Claims (3)

[Claims] (1) An image reduction method for reducing the size of a binary image, which has means for filtering the image, means for binarizing the filtering result, and means for subsampling for reduction, and has a method for reducing the size of the image. An image reduction device comprising means for referring to the periphery of a pixel and already reduced pixels, and correcting the binarization result for a specific pattern. (2) In claim (1), the reference means includes property detection means for detecting the properties of surrounding pixels including the pixel of interest, and a pattern of pixels around the pixel of interest and that have already been reduced. An image reduction device characterized by comprising a pattern detection section that detects a pattern. (3) The image reduction device according to claim 2, further comprising means for adaptively selecting a correction pattern based on the detection result of the property detection means. 4) The image reduction device according to claim 2, wherein the output of the filter or the coefficient of the filter is corrected based on the detection result of the property detection means.