JPS62221269A - Image conversion system - Google Patents

Abstract

PURPOSE:To comparatively easily, and accurately convert an image by converting an actual picture element by quantizing and patternizing the proportion that a picture element before conversion occupies in a picture element after the conversion when the image is enlarged or reduced. CONSTITUTION:Respective right-side, downward, and right downward picture elements are read from a picture memory as an attended picture element p (n, m) to before enlarging picture element, and set in a register 10. These four picture elements are selected by a selector 12 in accordance with the content of a ROM 18, and supplied to corresponding one of proportion quantization level arithmetic circuits 21, 22...2l. In the ROM 18, the pattern for the relevant picture element corresponding to a picture after enlargement is stored. This pattern is read out and inputted to the selector 12 in order to output the luminance levels of the four picture elements in the register 10 to one of the circuits 21-2l that corresponds to this pattern. As a result, the enlargement and reduction of a picture is executed by collecting by the prescribed amounts of data from upward, downward, right and left picture elements, and the deterioration of the picture is reduced.

Description

[Detailed Description of the Invention] [1] When reducing or enlarging an image, the proportion of pixels before conversion to the original pixels to be converted is quantized and patterned, and the actual pixels can be converted relatively easily. Perform accurate image conversion.

[Industrial application field]

The present invention relates to an image conversion method for enlarging or reducing an image represented by a group of pixels.

[Conventional technology]

In facsimiles, images are represented by groups of pixels.

For example, the number of scanning lines is 8 mm, and the number of samples of 1 scanning line data is also 8 mm, so the number of dots (pixels) in 1 dragon 2 is 64.
An image is represented by a group of pixels like this. The brightness (density) level of each pixel obtained by scanning the original image is an analog quantity, but it is converted into 4 bits and 16 bits at the sampling stage.
Make it into stages. After signal processing, this is converted into binary data of 1.0, and this is used for recording after transmission.

The manuscript size is A3. A4. B4, etc., and there is a request to reduce an A3 document to an A4 document and transmit it, or conversely, to enlarge it and transmit it.

[Problem that the invention seeks to solve]

Conventionally, a method of thinning out pixels has been generally used to reduce screen size. For example, the number of dots in one scanning line of a B4 original is 20.
4811M, A4 manuscript has 1728 pieces, 6
If you remove pixels at a ratio of IlI or 1 in 7, it becomes B4
You can reduce the original to A4 original.

However, with this method, the image quality deteriorates and it is not possible to obtain a high-quality screen as in the case of reducing the size using an optical system.

The present invention attempts to reduce or enlarge the screen without significantly deteriorating the image quality, and without using an optical system and only by data processing.

[Means for solving problems]

As shown in FIG. 2, the present invention patterns the reduction weighting. As shown in Figure 3, the reduction is made by solid line small rectangle A,
I2. ...A21, A22, ...
The chain line small rectangle Bll, B10. ...B11°B22. This is done by enlarging it to...

That is, solid line small rectangles All, Al1. ...is 84
The number of pixels on one scanning line of the original is 20481[1j
Yes, this is a chain line small rectangle Bll, B10.・・・・・・
If you enlarge the dots to 1728 on the same length and operate the printing (tIiii) ta mechanism, the printing mechanism is the same (the pitch of the dots remains unchanged), so (line 17)
One line of 28-dot A4 size image is completed.

This figure 3 can be thought of as an A4 original stacked on top of a B4 original stack, with the A4 original enlarged so that both are the same size.For this reason, the pixels of the A4 original are enlarged to 1728 pixels.
The length is the same as 2048 pieces of B4, and the scanning line (
There are also 1728 rows (horizontal rows), the same length and vertical width as B4's 2048 rows. Therefore, if you print small rectangles with chain lines, the number of scanning lines will be 1728, and both vertical and horizontal B4-A
A reduction of 4 is made.

When a solid line small rectangle is enlarged to a chain line small rectangle, the small rectangle after enlargement is made up of a plurality of small rectangles before enlargement. For example, Bll is A
All of ll and Al1. A21.

A22 consists of each part, and B12 consists of each part of Al1. A13゜A
22. Consists of each part of A23. The proportion of the small rectangles before enlargement and the small rectangles after enlargement varies slightly, but if this is quantized, the 16 patterns shown in FIG. 2 can be obtained. The pattern (0,0) has a large contribution from the pixel Aij, and its horizontal pixel Δi j+1 and the lower A I+ I J
A case is shown in which the contribution of the diagonally downward pixels A, . In FIG. 3, the small rectangle Bll corresponds to this. Pattern (0,1) is Aij and Aij+1
The contribution of is large and similar, and A 1+14 and 8i
This is a case where the contribution of ll j+1 is small and of the same degree, and this corresponds to the small rectangle B23 in FIG. Similarly, the enlarged rectangle shown in FIG. 3 can be made to resemble any of the 16 patterns shown in FIG. 2.

The pattern shown in FIG. 2 is obtained by quantizing the enlarged rectangle by dividing it into four parts vertically and horizontally, and dividing it into 16 parts in total.The pattern shown in FIG. 3 was obtained by quantizing the pattern shown in FIG. The pattern that appears is repetitive, for example B10 is B1
The pattern is similar to 1. Similar repeatability can be seen not only in the horizontal direction but also in the vertical direction. Also, when quantized as shown in Figure 2, the contributing small rectangles before expansion are either 4, 2 (contractor) or 1 (contractor).Pattern (3
, 3) is one unique case, which can be seen in the small rectangle B66 in FIG.

The brightness level of each pixel before enlargement is 4 bits and has 16 steps, but by weighting and adding them according to the contribution rate and taking the average, the brightness level of the pixel after enlargement can be obtained. The contribution rate is determined by the position of the pixel without actually measuring it, and any of the patterns in Figure 2 can be assigned, so the contribution rate of each pixel before enlargement for each pattern in Figure 2 can be determined by the position of the pixel. If , the brightness level of the enlarged pixel can be immediately calculated using it. For example, for pattern (0,0), as is clear from the drawing, the contribution rate of Aij is 9, the contribution rate of A i j+ l and A, the +th eye is 3, and the contribution rate of A, is l, so
lJ+1 The brightness level Bij of the pixel at turn (0,0) is Bij=
(9Aij+3A,, +3A, +A
)/IJ+1 ++1j i+
It becomes 1j+116. Note that pixels and their luminance levels are indicated by the same symbols here.

The pattern that a pixel takes after enlargement can be determined by its position. For example, Bll and B10 have a pattern (0,0).
, B13 is pattern (0,1), B10 and B10 are pattern (0,2), B,6 is pattern (0,3),
Grave B 7 is the same as B11. 2nd line B2+, B22. The same goes for .
2),... The following shall apply accordingly. Therefore, it is possible to prepare a counter, use the counter to count the clocks corresponding to the pixels after enlargement, generate a pixel number, use this as an address to access the ROM, and obtain the pattern number and thus the contribution rate of each pixel before enlargement. It is.

〔Example〕

An embodiment of the present invention is shown in FIG.
, m), the pixel P (n, re÷1) on the right, the pixel P (n+1+ m) below, the pixel P (diagonally below the right)
n+1.

4 pixels of m+1) are read out and these are set.

This register 10 may be associated with a window used for scanning on an image memory. 12 and 14 are selectors, 16 is the aforementioned counter, and 18 is a ROM (read-only memory). 21, 22゜...21 is for ratio quantization level 1, ratio 2, . . . for each arithmetic circuit,
These circuits 21.22. . . . adds input pixels with weights according to the patterns (0, 0), (0, 1), . . . and takes an average.

To explain the operation, the clock n of the pixel after enlargement is input to the counter 16, and the count value of the counter is reset to 0.
From the state of 1.2. ...... and so on. The post-expansion prime number is 1728 as described above, and the counter 16 can simply be set to 1728 base, and the signal m is the overflow pulse of the counter counted by a counter (not shown) (the number of scanning lines is the same 1728 as shown in the figure). 1728), the ROM 18 is accessed using this signal m and the count value of the counter 16 as an address, and the ROM 1B stores the pattern for each pixel after enlargement. If it is set, the pattern of the pixel can be read from the ROM 18. The output of this ROM18 is the selector 12
inputs the four pixel brightness levels of the register IO to the arithmetic circuit for the read pattern, and also inputs the brightness levels of the four pixels in the register IO to the selector 14
and outputs the output of the arithmetic circuit as the expanded R element luminance level.

For example, m=1. If the count value of the counter 16 is also 1, this means that pixel B is the calculation target, and the RO
M1B outputs a pattern (0, O). This causes the selector 12 to contain the contents of the register 10 Az, Al1. A2+
, A22 are input to the arithmetic circuit 21 for pattern (0,0), and the arithmetic circuit inputs (9A1.+3A, 2 +3A2.
+A22)/16=Bz is calculated, and the selector 14 outputs the calculated value B11 of this arithmetic circuit.

Next, one clock n enters and the count value of counter 16 is 2.
This and m=1 serve as an address to access the ROM 18 and output the pattern (0, O). Also, the contents of register IO are shifted one place to the right on the image memory and A12. A13゜A22. It becomes A23. Therefore, the selector 12 selects the calculation circuit 21 again, but the data to be calculated is different, and the calculation result is B12 = (9A12
+3A13 +3A22 +A23 )/16. This B10 is outputted by the selector 14. When the clock n enters again next time, the count value of the counter 16 becomes 3, m becomes 1, the output of the ROM 18 becomes the pattern (0, 1), and the contents of the register 10 become Al1. ．． A14゜A23. A2
4, and the arithmetic circuit 22 calculates 813=(6A13 +6A
+4 +2A23 +2A24 )/16, and the selector 14 outputs the BI3. The same applies below.

By utilizing the repeatability of the pattern, the counter 16 can be a decimal counter up to one cycle, and accordingly, the number of addresses in the ROM 18 may be small.

Above, we have mentioned the case of reducing B4 to A4, but A3
Other reduction examples, such as reducing the size of the image to A4 size, can be processed in the same way. The reverse enlargement, such as increasing A4 to 84, can also be processed in the same manner. That is, in this enlargement, the chain line small rectangle B11.8
+2. ...B21°B22.・・・・・・A
4 pixels, solid line Kochi-shape A11°A12.・・・・・・A
21. 822. . . . may be taken as a B4 pixel. If this conversion from Bij to Aij is performed, the number of Bij on one scanning line is 1728, the number of Aij is 2048, and the printing (drawing) mechanism remains unchanged, so the A4 original is enlarged to a B4 original. The calculation procedure in this case is All = Bll
, Al1 = (3B11+9812) /12.
A13 = (3812+9813)/12, Al1
= (B 13 +814) /2.・・・・・・
It is clear from FIG.

〔Effect of the invention〕

The enlargement and reduction of the present invention does not involve thinning out pixels to reduce the size or adding the same @ element to enlarge the image, but instead of increasing or reducing pixels by adding a predetermined amount from the upper, lower, left, and right pixels, there is an advantage that there is little deterioration in image quality. The amount by which the pixels on the upper, lower, left, and right sides are brought together is determined by the position of the pixel, and can be prepared in advance, so it is easy to calculate the luminance (density) of the converted pixel. The amount (contribution rate) of the top, bottom, left, and right pixels can be quantized and patterned. Since this pattern has repeatability, a relatively small number of patterns such as 16 patterns is sufficient, and the number of arithmetic circuits and RO
M capacity etc. can be reduced and simplified.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing an embodiment of the present invention, FIG. 2 is an explanatory diagram of a reduction weighting pattern, and FIG. 3 is an explanatory diagram of screen reduction/enlargement. In FIGS. 1 and 3, All, A12. ...Bl
l. B12. . . . P(n, n+) . . . is a pixel or its brightness level, and 18 is a ROM that outputs a pixel pattern. 21.22. . . . is a circuit that performs calculations according to each contribution rate for each pattern.

Claims (1)

[Claims] In an image conversion method that enlarges or reduces an image represented by pixels by decreasing the number of pixels, the converted pixels (B_1_1) obtained by overlapping the pre-conversion and post-conversion screens assuming the same screen size. Multiple pixels before conversion (A_1_1, A_1_2, A_2_1, A_2_2
) is quantized and patterned, and the brightness level of the pixel after conversion is calculated and determined from the brightness level of the pixel before conversion according to the pattern and the contribution rate.