JPH04222166A - Image smoothing device - Google Patents

Abstract

PURPOSE:To obtain satisfactory images and to process images having the various kinds of resolution by transforming a rectangular pixel data to two first squares having the shorter side as one side and further transforming the first squares to four second squares so as to execute the two-stage processing. CONSTITUTION:A DMA read part 2 reads out the almost double pixel data, which one picture element is rectangule, having the almost double side as long as the shorter one and when inputting such a pixel data, this is transformed to the pixel data for two square picture elements in the first size having the shorter side as one side. Afterwards, each pixel data in the first side is transformed to the pixel data for four square picture elements in the second size having the further half length of the shorter side as one side. This two-stage transformation processing is executed by a latch part 10, data segment part 11, first and second picture element correction parts 1201 and 1202. On the other hand, when inputting the pixel data in the first size, the second-stage processing is directly executed. Thus, the satisfactorily corrected images can be obtained by simple picture element correction algorithm, and images having the various kinds of resolution can be processed.

Description

[Detailed description of the invention]

0001

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image smoothing device for smoothly correcting the contours of a binary image.

0002

BACKGROUND OF THE INVENTION The resolution of an image in a normal facsimile machine such as G3 standard is 8 dots per mm x 3.
It is set to one of 85 lines, 8 dots x 7.7 lines, 8 dots x 15.4 lines, or 16 dots x 15.4 lines.

[0003] Generally, an operation mode in which image processing is performed using 8 dots x 3.85 lines is called a standard (hereinafter abbreviated as STD) mode. Similarly, 8 dots x 7.7 lines is Detail (hereinafter abbreviated as DTL) mode, 8 dots x 15.4 lines is Semi-Super Fine (hereinafter abbreviated as SSF) mode, and 16 dots x 15.4 lines is Fine mode. (hereinafter referred to as FINE) mode.

By the way, in facsimile machines, when the recording resolution of the recording unit is higher than the resolution of the received original image, it is often done to perform smoothing processing on the original image and record the image. It is.

[0005] For example, suppose that a facsimile machine equipped with a FINE mode recording unit receives an image having a 45-degree slope as shown in FIG. 10(a) in DTL mode.

[0006] The facsimile machine executes smoothing processing on the original image, and as shown in FIG.
Image correction is performed to remove the step-like corners of the image and fill in the valleys. In this case, the algorithm for correcting each pixel of the original image is known.

Next, suppose that this facsimile machine receives a similar image in STD mode.

The width of one pixel in the STD mode in the sub-scanning direction is approximately twice that in the main scanning direction. Therefore, each pixel of the received image has a vertically elongated shape, as shown in FIG. 11(a).

When smoothing processing is performed on this received image using the above-mentioned known algorithm, the corners and valleys of the image are not completely corrected, as shown in FIG. 2(b). Furthermore, since each pixel in the corrected portion also has a vertically elongated shape, a smooth and natural image cannot be obtained.

[0010] Therefore, in order to correctly correct the received image in the STD mode, smoothing processing must be performed using a new algorithm exclusively for the STD mode.

Next, suppose that the facsimile machine receives a similar image in SSF mode.

In the SSF mode, one pixel of the received image has a horizontally long shape. Therefore, as in the case above,
In order to correct the received image correctly, smoothing processing must be performed using yet another algorithm dedicated to the SSF mode.

[0013]

[Problem to be solved by the invention] In this way, conventionally,
If the pixel shape of the original image is not square, a good corrected image cannot be obtained using a known simple algorithm, and smoothing processing must be performed using a different algorithm for each resolution. For this reason, there has been a problem that the processing circuit for smoothing processing has to be specialized or the number of circuits has increased, resulting in an increase in device cost.

SUMMARY OF THE INVENTION An object of the present invention is to solve the above problems and provide an image smoothing device that can obtain good images with a simple algorithm and can process images of various resolutions with a small number of circuits. do.

[0015]

[Means for Solving the Problems] To this end, the present invention
When inputting pixel data where one pixel is a rectangle and the long side is approximately twice the short side, convert the input pixel data to pixel data for two square pixels of the first size with the short side as one side. Then, in a two-step procedure, each pixel data of the first size is converted into pixel data of four pixels of a second size square whose sides are half of the short side. I'm trying to process it.

Furthermore, when pixel data corresponding to the first size is input, the input pixel data is directly processed into four pixels of the second size by the second stage processing means. I am trying to convert it to data.

Furthermore, in the third invention, when inputting pixel data of an original image in which one pixel is a rectangle and the long side corresponds to one side of the first size, the inputted pixel data is
The pixel data is processed by the second stage processing means, and pixel data for two pixels in the long side direction are extracted from among the four pixels obtained by the second stage processing means.

[0018]

[Operation] When processing pixel data where each pixel is a rectangle and the long side is approximately twice the short side, such as an image in STD mode, the first stage processing algorithm is relatively simple.
A known algorithm can be used as is for the second stage algorithm. This makes it possible to obtain good images using a simple algorithm.

Furthermore, when pixel data corresponding to the first size is input, as in the case of an image in DTL mode, the second stage processing is executed as is, while, as in the case of an image in SSF mode, If you input pixel data of an original image in which the pixels are rectangular and the long side corresponds to one side of the first size, the second stage processing described above is executed and only two of the four pixels are input. By extracting the pixel data, predetermined pixel data is obtained, so that not only a good image can be obtained, but also images of various resolutions can be processed with a small number of circuits.

[0020]

DESCRIPTION OF THE PREFERRED EMBODIMENTS Hereinafter, embodiments of the present invention will be described in detail with reference to the accompanying drawings.

FIG. 1 shows a block diagram of an image processing apparatus according to an embodiment of the present invention. In the figure, a page memory 1 temporarily stores input pixel data for one page of images. The DMA reading unit 2 reads pixel data from the page memory 1 using DMA (Direct).
The information is sequentially read out using Memory Access) control.

The smoothing processing unit 3 performs smoothing processing on the input image by correcting the read pixel data using a certain method.

The recording unit 4 is, for example, an image recording device using a laser recording method, and records the smoothed image using an FI.
The recording paper is sequentially recorded line by line at a resolution of 16 dots x 15.4 lines, which corresponds to the NE mode. The control section 5 monitors and controls each of the above sections.

FIG. 2 shows a circuit diagram of the smoothing processing section 3. As shown in FIG. In the figure, the latch section 10 is
Four lines of pixel data are temporarily held one byte at a time.

The pixel data PD output from the DMA reading section 2 is stored in four registers 1001 to 1 in the latch section 10.
004 in parallel. Pixel data PD is a 1-byte parallel signal, and is stored in registers 1001 to 1001.
1004 each holds 1 byte of data. Each register 1001 to 1004 receives a clock signal CKi, CKi+1 to CKi from the DMA reading section 2.
+3 is input, thereby setting the pixel data of each line.

The data cutting section 11 cuts out four lines of pixel data in units of 3 bits.

The output of the register 1001 is sent to the P/S converter (parallel/serial) 11 in the data extraction section 11.
01 is input. The P/S converter 1101 outputs input 8-bit pixel data bit by bit as a serial signal. The output is input to an F/F (flip-flop) 1102, and the F/F 1102
The output of is input to F/F 1103.

Similarly to the above, the output data of the register 1002 is transmitted from the P/S converter 1104 to the F/Fs 1105 and 1.
106, the output data of the register 1003 is sent from the P/S converter 1107 to F/Fs 1108 and 1109, and the output data of the register 1004 is sent to the P/S converter 1110.
The data is sequentially transferred from the F/F 1111 to the F/F 1112, respectively.

[0029] P/S converters 1101, 1104, 110
7 and 1110, pixel data is set one byte at a time by a load signal L. In addition, those P/S converters and F/Fs 1102, 1103, 1
105, 1106, 1108, 1109, 1111 are
Pixel data is sequentially shifted and output one bit at a time in response to a clock signal CK1. Then, each output is processed as cut-out pixel data A to L by the data cut-out unit 11.
It is output from.

The pixel correction section 12 corrects the input pixel data using a certain algorithm, and includes two smoothing logic circuits 1201 and 1202. The smoothing logic circuit 1201 receives the cutout pixel data A to I as input and outputs two pixel data EU and ED corresponding to the pixel data E. The smoothing logic circuit 1202 uses the extracted pixel data D to L.
is input, and two pixel data HU and HD corresponding to pixel data H are output.

The signal line of the pixel data EU is connected to the contact a1 of the selector 13, and the pixel data ED is connected to the contacts b1 and a2. In addition, pixel data HU
are connected to contacts b2 and a3, and pixel data HD is connected to contact b3. Common terminals c1 to c3 of the selector 13 are connected to the data extraction section 14.

The data cutting section 14 cuts out three bits of input pixel data of three lines. The signal lines for the pixel data E, H, and K are connected to the contacts a4 to a6 of the selector 1401 in the data cutting section 14. Further, common terminals c1 to c3 of the selector 13 are connected to the contacts b1 to b6, respectively.

Output data from the common terminal c4 of the selector 1401 is sequentially transferred to F/Fs 1402 and 1403. Similarly, output data from the common terminal c5 is sequentially transmitted to F/Fs 1404 and 1405, and output data from the common terminal c6 is sequentially transmitted to F/Fs 1406 and 1407.

The F/Fs 1402 to 1407 sequentially shift and output pixel data in response to a clock signal CK2. And common terminals c4 to c6 and F
The respective outputs of /F1402 to 1407 are outputted from the data cutout section 14 as cutout pixel data a to i.

The pixel correction unit 15 uses a smoothing logic circuit 1501 to correct input pixel data using a certain method.

The smoothing logic circuit 1501 inputs the cut-out pixel data a to i and outputs four pixel data e0 to e3 corresponding to the pixel data e.

The selector 16 selects each pixel data e0 to e3.
are selected and output in sequence.

With the above configuration, the operation of the image processing apparatus of this embodiment when processing an image in STD mode will be explained next.

In this case, STD mode pixel data is sequentially input from another device. The page memory 1 temporarily stores the pixel data. Now, the pixel data is A
If the original image is 4 sizes, one line is 17
It is 28 bits or 216 bytes, which is data for 1143 lines. Such pixel data is sequentially stored one byte at each address of the page memory 1.

When the page memory 1 stores one page worth of pixel data, the DMA reading section 2 starts reading out the pixel data from the page memory 1.

In this case, the DMA reading section 2 first reads the first
The pixel data of the first byte at the beginning of each line from the line to the fourth line is sequentially read out and transferred to the smoothing processing section 3. Next, the pixel data of each line is sequentially transferred one byte at a time until the 216th byte at the end of the line, such as transferring the second byte of each line in the same way.

[0042] When the transfer operation from the first line to the fourth line is completed, the same transfer operation is repeated three more times, for a total of 4 lines.
Execute repeatedly.

[0043] When the four transfer operations are completed, next
Four lines of pixel data from the second line to the fifth line are transferred in the same way.

In this way, the pixel data of lines i to i+3 are sequentially transferred to the smoothing processing section 3 four times each.

The smoothing processing section 3 executes smoothing processing on the pixel data.

FIG. 4 shows the smoothing process. That is, the smoothing processing unit 3
When processing pixel data of mode, selector 1401
is switched to contact b side (process 2001), and selector 13
is switched to the contact a side (processing 2002).

Each pixel data of line i to line i+3 sequentially transferred from the DMA reading section 2 is stored in the register 1001.
~1004, and then simultaneously transferred to P/S converters 1101, 1104, 1107, and 1110. And each P/S converter 1101, 1104, 110
7, 1110 are sequentially output one bit at a time, and each pixel data is output from each F/F 1102, 1103, 110.
5,1106,1108,1109,1111,111
It is held at 2. As a result, pixel data A to L for 4 lines x 3 pixels are cut out (processing 2003).

As shown in FIG. 5A, the smoothing logic circuit 1201 inputs the pixel data A to I for the first three lines of the four lines and outputs two pixel data EU and ED. do. In addition, as shown in FIG. 2B, the smoothing logic circuit 1202 inputs the pixel data D to L for the last three lines of the four lines, and
Outputs two pixel data HU and HD.

By the way, the resolution in the STD mode is 8 dots/mm in the main scanning direction and 3.85 lines/m in the sub-scanning direction.
m, one pixel is approximately twice as large in the vertical direction as in the horizontal direction.

[0050] Smoothing logic circuits 1201 and 1202 focus on one pixel in the center among the nine input pixels, and output two upper and lower pixel data EU, ED and HU, HD according to each surrounding pixel pattern. do.

That is, FIGS. 6A to 6L show the algorithm for this pixel conversion. Hatching in the figure indicates black pixels, and diagonal lines indicate that neither black pixels nor white pixels affect the output.

45 in the image.
This is a pixel conversion pattern that detects one pixel in a tilted portion of the angle and decomposes it into a white pixel and a black pixel. Figures (e) to (h) show pixel conversion patterns for removing isolated points above the contour of the image. Further, (i) to (e) of the same figure are pixel conversion patterns for removing isolated points below the contour of the image.

In this way, the smoothing-corrected 8 dots x 7.7 lines, that is, the pixel data E in the DTL mode.
U, ED, HU, and HD are obtained (process 2004).

For example, if STD mode pixel data of a diagonal line image as shown in FIG. 7(a) is input, the pixels are corrected by the smoothing process as shown in FIG. 7(b). It turns out.

Since the selector 13 is initially connected to the contact a side, the pixel data EU, ED, HU are sequentially input bit by bit to the data cutting section 14, and are divided into 3×3
Bit pixel data a to i are extracted (processing 2
005).

Next, as shown in FIG. 5(c), the pixel correction unit 15 inputs the above pixel data a to i, and divides the central pixel data e into four pixels corresponding to the top, bottom, left, and right of the pixel data. Convert to pixel data e0 to e3.

FIGS. 8(a) to 8(r) show the pixel conversion algorithm in this case. That is, as is clear from the figure, pixel correction is performed to shave off the corners and fill in the valleys in the shaded areas of the input image. This results in
The 45 degree inclined portion detected by the pixel correction unit 12 is now further smoothly corrected.

In this way, smoothing-corrected pixel data with a resolution of 16 dots x 15.4 lines, ie, FINE mode pixel data is obtained (processing 2006).

As a result, the image shown in FIG. 7(b) is corrected into a smoother image as shown in FIG. 7(c).

By the way, as described above, pixel data at the same line position is read out from the page memory 1 four times at a time, and in response to one reading operation, recording scanning of one line is performed in the recording section 4. That's what I do. Therefore, recording scanning is also executed four times for each read line position of the page memory 1.

[0061] Once the above corrected pixel data is obtained,
It is determined how many times the scan is currently being performed (processing 2007).

Here, for example, if it is the first scan ("1" in process 2007), the selector 16 sequentially selects and outputs pixel data e0 and e1 (process 200).
8).

After this, the position on the line of the pixel data currently being read from the page memory 1 is determined (processing 2
009). If it is not the end of the line (processing 200
(N of 9), the same process is repeated until the end of the line while shifting the pixel data of each line one bit at a time into the data cutting unit 11 (proceed to process 2003).

At this time, the selector 16 selects the pixel correction unit 1
5 pixel data e0 and e1 are sequentially and alternately taken out. As a result, one line of pixel data in FINE mode is obtained. This one line of pixel data is input to the recording section 4. The recording unit 4 performs one recording scan at a resolution of 16 dots/mm to record one line of image.

[0065] When the end of the line is processed in this manner (Y in process 2009), it is determined which scan is the end (process 2010). Now, if it is the first scan (from N in process 2010, process 2011, process 201
1N), then a second reading of pixel data at the same line position from the page memory 1 is started, and the same process is executed (proceed to process 2003).

In the case of this second scan (processing 200
7 "2"), the selector 16 selectively outputs the pixel data e2 and e3 of the pixel correction unit 15 (process 2012).

Upon completion of this second scan (processing 201
0 Y), switch the selector 13 to the contact b side (process 2013), and repeat the same process (process 200
3).

In this case, by switching the selector 13, the data from the pixel correction section 12 to the data extraction section 14 is transferred from the pixel correction section 12 to the data extraction section 14 as shown in FIG.
), pixel data ED, HU, HD are input, and based on these pixel data, the pixel correction unit 15
Generate pixel data e0 to e3 (processing 2006)
.

In the case of this third scan (processing 200
7 "3"), and outputs pixel data e0 and e1 (processing 2008).

After this, the fourth scan process is executed in the same manner. In the case of the 4th scan ("4" of process 2007)
), and outputs the pixel data e2 and e3 of the pixel correction unit 15 (processing 2012).

[0071] The recording section 4 records the image of the four lines of pixel data thus output while performing sub-scanning at a rate of 15.4 lines/mm. As a result, the input 1-line image in STD mode is recorded as a 4-line image in FINE mode.

When recording of 4 lines for 1 line in this STD mode, that is, 4 scans is completed (processing 20
11 Y), determine the presence or absence of the next line (process 201
4). If there is a next line (N in process 2014), the read address of the page memory 1 is advanced by one line (process 2015), and the same process is repeated (proceed to process 2002).

In this way, the pixel data of each line of the page memory 1 is sequentially processed, and when all lines have been processed (Y in process 2014), the above STD mode image processing is completed.

As described above, in this embodiment, the pixel data in the STD mode is corrected to the pixel data in the DTL mode by detecting the 45 degree diagonal portion of the input image in the pixel correction section 12, and then, The pixel correction unit 15 corrects the pixel data to FINE mode pixel data using a known algorithm.

[0075] Through such two-stage processing, a smooth and natural image can be obtained. Furthermore, in this case, the pixel correction unit 1
Algorithm 2 is simple, and the pixel correction unit 15
A known algorithm can also be used.

Next, the image processing apparatus of this embodiment uses DTL
The operation when processing a mode image will be explained.

Pixel data in the DTL mode is input from an external device and temporarily stored in the page memory 1, as described above.

In this case, the DMA reading section 2 first sequentially transfers the pixel data of the first line to the register 1002 of the latch section 10, the second line to the register 1003, and the third line to the register 1004, one byte each. do. This transfer operation is repeated twice.

Thereafter, similar transfer operations are performed from the second line to the fourth line. In this way, the pixel data is transferred to the latch unit 10 three lines at a time.

In this case, the smoothing processing unit 3
As shown in FIG. 3, first, the selector 1401 is switched to the contact a side (processing 3001). At this time, the input 3
The pixel data of the line is sequentially shifted into the data cutting section 11, and from the data cutting section 11, the pixel data D~
L is cut out.

The data cutting unit 14 sequentially shifts in the pixel data E, H, K and cuts them out as pixel data a to i for 3 lines x 3 bits (processing 3002).
.

[0082] The pixel correction unit 15 uses the pixel data a to i.
Input , correct the center pixel in the same way as above, and get 4
Output pixel data e0 to e3 (processing 3003)
.

When processing pixel data in the DTL mode, the recording section 4 performs recording scanning twice for each line.

Here, it is determined whether the current process corresponds to the number of recording scans (process 3004), and if it is the first scan ("1" in process 3005), the selector 16 selects the pixel data e0. and e1 are selectively output (processing 3005).

This operation is repeated in the same way (process 3006
(Proceed to process 3002 from N). As a result, the input 1
By processing the pixel data of a line once until the end of the line, pixel data for one line is output to the recording section 4. During this time, the recording unit 4 executes one recording scan and prints 16
One line image is recorded with a resolution of dots/mm.

Then, when the second scan operation starts ("2" in process 3004), the pixel data e2, e3
is selected and output (processing 3007).

The recording section 4 records the pixel data of each line thus output as an image at a resolution of 15.4 lines/mm. As a result, the input one-line image in DTL mode is recorded as a two-line image in FINE mode.

When the above operations for the second scan are completed (Y in process 3006), the presence or absence of the next line is checked (process 3008). Then, if there is the next line (processing 3
008 N), advances the read line of page memory 1 by one line (process 3009), and repeats the same process (
(Proceed to process 3002).

When the last line has been processed (Y in process 3008), the above DTL mode image processing is completed.

In this way, since the pixel data in the DTL mode undergoes smoothing processing by the pixel correction section 15, the processing circuit can be shared with that in the STD mode.

Next, the image processing apparatus of this embodiment uses SSF
The operation when processing a mode image will be explained.

Pixel data in the SSF mode is input from an external device as described above and is temporarily stored in the page memory 1.

The DMA readout section 2 reads pixel data for three lines from the page memory 1 and transfers it one byte at a time to the smoothing processing section 3, as in the case of the DTL mode, but in this case, the transfer operation is Execute once.

In the smoothing processing unit 3, as shown in FIG. 10, the selector 1401 is switched to the contact a side (processing 4001), and the data cutting unit 14 converts the pixel data of the three lines into three lines. Pixel data a to i for line x 3 bits are extracted (processing 4002).

The pixel correction unit 15 uses the pixel data a to i
A smoothing process similar to that described above is performed on the image (process 4003).

By the way, one pixel of an image in SSF mode has two pixels in the main scanning direction compared to one pixel in FINE mode.
It is twice the size and has a horizontally elongated shape. Therefore, S.S.
If one pixel in F mode is divided into two pixels corresponding to its left and right sides, pixel data in FINE mode can be obtained.

In (a) to (r) of FIG. 8, it is necessary to convert the white or black pixel of interest into two pixels, white and black on the left and right, in (d), (g), (m), (q) 4
There are two cases. In addition, in other various conversion patterns, among the four pixel data e0 to e3 that are the conversion results, the lower two pixel data e2 and e3 each have the same color as the pixel of interest.

Therefore, in this case, the lower two pixel data e2 and e3 output from the pixel correction section 15 are sequentially extracted (processing 4004).

[0099] This makes it possible to obtain pixel data that has been smoothed and corrected from the SSF mode to the FINE mode.

This process is executed until the end of the line (from N in process 4005 to process 4002). In this case, when that one scan is completed (Y in process 4005), the read line of page memory 1 is immediately advanced by one line (process 4006, from process 4006 to process 4007), and the same process is repeated (process 4002). .

[0101] In this way, the pixel data in the SSF mode is processed by the pixel correction unit 15, and among the four output pixels,
By extracting the bottom two pixels, FINE mode pixel data is obtained.

[0102] Thereby, even in the SSF mode, the pixel correction section 15 can perform smoothing processing.

[0103] In the above embodiments, the case where the facsimile machine performs smoothing processing on recorded images has been explained as an example. It goes without saying that the invention can be similarly applied.

[0104]

As described above, according to the present invention, when inputting pixel data in which the long side of one pixel is approximately twice as long as the short side, the input pixel data is transferred to the first pixel data whose short side is one side. Smoothing is performed using a two-step procedure of converting into pixel data for two square pixels of the same size, and then converting it to pixel data of four pixels of a second size square with half of the shorter side as one side. Since the processing is performed, a good corrected image can be obtained using a simple pixel correction algorithm.

[0105] When inputting pixel data of the first size, the second stage processing is executed as is to obtain corrected pixel data, while one pixel is a rectangle with long sides of the first size. When inputting pixel data corresponding to one side of the size, corrected pixel data is obtained by executing the second stage processing and extracting only two pixels out of the four pixels obtained. Therefore, good corrected images can be obtained, and images of various resolutions can be processed with a small number of processing circuits.

[Brief explanation of the drawing]

FIG. 1 is a block configuration diagram of an image processing device according to an embodiment of the present invention.

[Figure 2] and

FIG. 3 is a circuit configuration diagram of a smoothing processing section.

FIG. 4 is an operational flowchart of smoothing processing for pixel data in STD mode.

[Fig. 5] Explanatory diagram showing the operation of the first stage pixel correction section

[Figure 6
] An explanatory diagram showing a pixel correction algorithm of a first stage pixel correction section.

FIG. 7 is an explanatory diagram showing an example of image correction in STD mode.

FIG. 8 is an explanatory diagram showing a pixel correction algorithm of a second stage pixel correction section.

FIG. 9 is an operation flowchart of smoothing processing for pixel data in DTL mode.

FIG. 10 is an operation flowchart of smoothing processing for pixel data in SSF mode.

FIG. 11 is an explanatory diagram showing a conventional example of image correction using a known algorithm.

[Figure 12] STD corrected by the above known algorithm
FIG. 2 is an explanatory diagram showing a conventional example of a mode image.

[Explanation of symbols]

1 Page memory 2 DMA reading section 3 Smoothing processing section 4 Recording section 5 Control section 10 Latch section 11, 14 Data extraction section 12, 15 Pixel correction section 13, 16, 1401 Selector 1201, 1202, 1501 Smoothing logic circuit

Claims (3)

[Claims] 1. An image smoothing device that smoothly corrects the outline of an image by converting pixel data at each pixel position in an original image of one page into an integer number of pixel data based on pixel patterns of adjacent pixels. , the pixel shape per pixel is rectangular, and the long side is about 2 times the short side.
a first pixel data input means for externally inputting pixel data of the original image, which is double the size of the original image; a first pixel correction means for converting the converted pixel data of the first size into pixel data of four square pixels of a second size having one half of the short side as one side; An image smoothing device comprising: second pixel correction means. 2. A second pixel data input means for directly inputting pixel data of the original image whose pixel shape per pixel corresponds to the first size; 2. The image smoothing apparatus according to claim 1, further comprising a control means for converting pixel data of four pixels of said second size by said pixel correction means. 3. Third pixel data input means for directly inputting pixel data of the original image from the outside, in which the pixel shape of each pixel is a rectangle and the long side corresponds to one side of the first size; control means for converting the pixel data into pixel data for four pixels of the second size by the second pixel correction means; 2. The image smoothing device according to claim 1, further comprising pixel extraction means for extracting pixel data.