JPH0423869B2 - - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION (a) Field of Industrial Application This invention relates to a method for compressing a digital image signal obtained by scanning a color original image using an image scanning recording device such as a color scanner.

(B) Prior art Color gradation images generally require rich gradation expression, and even when digitized, each of the three primary colors of light, red (R), green (G), and blue (B), or cyan ( It is said that each ink color of C), magenta (M), yellow (Y), and black (K) requires the ability to express approximately 200 or more gradations (8 bits).

In this way, in a color gradation image, the number of bits required per pixel is large, so for example, if the pixels in the scanning area are recorded in a memory device for each separated color over the entire scanning area, the memory device will have a large number of bits. The memory capacity becomes enormous. Therefore, in order to save the total memory capacity as much as possible, the sampling pitch when creating an image signal for color gradation images is increased (roughened) to the limit that does not feel unnatural, in line with human vision. Therefore, the correlation between each adjacent pixel cannot be said to be extremely high, and for example, it is difficult to compress the image signal for each color separated image of R, G, B or C, M, Y, K. Even so, it is usually difficult to increase the compression ratio without causing major problems when the image is played back.

Based on this problem, a method for recording color images that significantly reduces memory capacity and hardly impairs visual quality was proposed in Japanese Patent Application Laid-Open No. 55-22708 (Patent Application No. 53).
-94507) by the applicant. This method shows that while human vision is very sensitive to changes in brightness and darkness in minute areas, it is also very sensitive to changes in color in minute areas. Two-dimensional compression and reproduction of color images is performed by taking advantage of the fact that the color image is not

That is, at least one of the color-separated color image signals (generally, a value representing brightness) is left, other digital image signals at each pixel are omitted in a predetermined relationship, and the signal is stored in a memory device. The present invention relates to a compression method and a reproduction method. Below, C, M, Y, mentioned in the above-mentioned Japanese Patent Application Laid-Open No. 55-22708,
An example of compressing each K color version signal will be described. Figure 1 shows M, which is relatively close to the sensitivity of human clear vision among the C, M, Y, and K4 color signals.
The color plate signal is substituted for the brightness value, and the other C, Y, K color plate signals are provided only in the representative pixel of the compression unit area 2 x 2 pixels, and the other pixels are provided with the C, Y, K color plate signals. FIG. 7 is a schematic diagram in which all pixels are provided with only M color signals corresponding to brightness values by omitting each color signal.

The M color signal corresponding to the brightness value in this case is said to have 8-bit information for both the representative pixel and other pixels.

In addition, in this case, the calculation formula for reproducing the omitted C, Y, and K color signals for pixels other than the representative pixel is such that the subscripts (00), (01), (10), and (11) are for the representative pixel, respectively. For example, for the pixel position (01), C ₀₁ = C ₀₀ + M ₀₁ − M ₀₀ M ₀₁ = M ₀₁ = M ₀₀ + M ₀₁ − M ₀₀ Y ₀₁ = Y ₀₀ + M ₀₁ −M ₀₀ K ₀₁ =K ₀₀ +M ₀₁ −M ₀₀ . In such a reproduction method, changes in color can only be dealt with for each compressed unit area, but M ₀₁ − common to the above four formulas
Since M ₀₀ is the difference in brightness between both pixel positions, there is an advantage that changes in brightness can be handled in units of the smallest pixel.

In addition, in the above formula, when calculating C ₀₁ , Y ₀₁ , and K ₀₁ , the case where C ₀₀ , Y ₀₀ , and K ₀₀ of representative pixels are substituted is shown, but it is also described in the above-mentioned Japanese Patent Application Laid-Open No. 55-2708. Of course, it is also possible to obtain interpolated values based on the values of the representative pixels at the four corners of the vicinity of the pixel to be determined, and to use these interpolated values instead. However, since this point is not the gist of the present invention, detailed explanation will be omitted.

According to this method, color information is provided only for the representative pixel of each compression unit area, and is omitted for other pixels, thereby making it possible to significantly save memory capacity. Specifically, C, M, Y, K, which has 8-bit information for a compression unit area of 2 x 2 pixels.
To store all of the data, a memory capacity of 2 x 2 pixels x 4 colors = 16 bytes is required, but by omitting C, Y, and K image signals other than the representative pixels, 2x
The memory capacity is 2 pixels x brightness value + 3 colors = 7 bytes.

However, even with this method, the C, Y, and K version image signals are omitted to 1/4 in the above example (1/9 when the compression unit area is 3 x 3 pixels), but the brightness is reduced. Since the image signal of the M version still contains details of every pixel, the compression rate of the image signal of the C, M, Y, and K versions as a whole was still insufficient.

(c) Purpose of the present invention The present invention provides the "color image recording method" disclosed in the above-mentioned Japanese Unexamined Patent Publication No. 55-22708, in which the brightness values are detailed in all pixels, even though the memory capacity can be significantly reduced. This was done to further improve the problem that the compression ratio as a whole was insufficient due to the need to maintain a compression unit of It is an object of the present invention to provide an image signal compression method capable of reducing the average number of bits per pixel of a signal responsible for brightness within a region.

(d) Structure of the present invention The image signal compression method according to the present invention is such that the signal responsible for brightness (hereinafter referred to as the brightness value) is not omitted from the color-separated image signal, and other signals are maintained in a predetermined relationship. It is constructed based on the premise of a compression method that is often omitted, and although the sampling pitch is set as large as possible to match human vision without causing an unnatural impression, it is possible to Focusing on the fact that there is a strong correlation (redundancy) between the brightness values in The gist is to encode the brightness values of pixels other than the representative pixel of the compression unit area by performing non-linear quantization based on the difference between the brightness values of the representative pixels of the compression unit area and generate a compressed image signal. It is something to do.

In other words, there is no sudden change in brightness between adjacent pixels, and even if there is a sudden change, the image before compression and the image after compression reproduction are not required to be exactly the same, so This reduces the total number of bits allocated to the brightness value.

(E) Embodiments An embodiment of the image signal compression method according to the present invention will be described below with reference to the drawings. As described above, this invention focuses on the redundancy between a plurality of brightness values within a unit area to be compressed, and compresses brightness values as well. This is a diagram schematically showing the state of an image signal compressed based on the method. In the figure, representative pixels C, M,
As for Y and K, as shown in FIG. 1, they are the same as in the conventional method. Also, as in Figure 1, the subscript (00),
(01), (10), and (11) each indicate the positional relationship with the representative pixel, and for other than the representative pixel, the difference between the brightness value of each pixel and the brightness value of the representative pixel is calculated, and a nonlinear quantization characteristic is applied to this. is compressed by giving . That is, m ₀₁ =M ₀₁ −M ₀₀ , m ₁₀ =M ₁₀ −
First, a difference signal is obtained for each pixel with M ₀₀ , m ₁₁ =M ₁₁ −M ₀₀ . These correspond to prediction errors, and if there is redundancy among M ₀₀ , M ₀₁ , M ₁₀ , and M ₁₁ , m 01 , m ₁₀ , _{and m 11 often} take extremely small values. Therefore, m _{′ 01} , m′ ₁₀ , m′ _{01 ,} m′ ₁₀ ,
It is compressed to the shape of m′ ₁₁ .

In such cases, variable-length encoding is often used, but here we consider cutting out the original image from an arbitrary position to be a very important element, so we will use fixed-length encoding as an example. Let me explain.

FIG. 3 is a diagram showing the conversion characteristics for nonlinear quantization. When the prediction error signals m ₀₁ , m ₁₀ , m ₁₁ are close to 0, they are fine-grained, and as they move away from 0, the representative value m becomes coarser. ′ ₀₁ , m′ ₁₀ , m′ ₁₁ are prepared. At this time, the total number of steps of the representative value is much smaller than the overall gradation value, and the number of bits can be reduced accordingly.

Reproduction when such nonlinear quantization is adopted will be explained based on FIG. 4. In the part A where the gradation changes gradually in the figure, the prediction error is small, so the prepared representative values can be reproduced finely and accurately. On the other hand, in part B where the gradation changes rapidly, the prediction error becomes large, so the deviation from the actual representative value may become large, and in that case, the reproduced duplicate image may be slightly different from the original image. become. Furthermore, if the prediction error becomes too large, there are cases where a representative value is not prepared, and the deviation in that case becomes even larger. However, in the diagram of the nonlinear quantization characteristics shown in Figure 3, the representative value m′ ₀₁ and
m' ₁₀ is about 5 bits (32 levels), and the representative value m' ₁₁ is about 6 bits (64 levels). At this time, m' ₀₁ ,
For m′ ₁₀ , when the absolute value of the prediction error is extremely large, the representative value is truncated at a constant value, but compared to the position (11) in the compression unit area, the position (01), (10)
Since the position is close to the representative pixel position (0.0), such a situation will not occur.

Therefore, the brightness value M ₀₁ of the pixels other than the representative pixel,
In the uncompressed form, M ₁₀ and M ₁₁ each require 8 bits and a total of 3 bytes, but with this encoding, m' ₀₁ (5 bits), m' ₁₀ (5 bits), m' It is compressed to ₁₁ (6 bits) and only takes 2 bytes.

Although the above-mentioned embodiment describes the case of four-color signals of C, M, Y, and K, the method according to the present invention is not limited to such color signals; It can also be applied to three-color separation signals,
In such a case, the G color separation signal is used as the signal responsible for brightness.

In addition, in the above embodiment, the case where the compression unit area is 2×2 pixels is described, but 3×
Of course, it is also possible to use 3 pixels, 4×4 pixels, etc.

FIG. 5 is a block diagram when the method according to the present invention is applied to, for example, a layout scanner system, in which a type of system in which the original image scanning section 1 and the recording section 2 are separated is shown. The original image scanning unit 1 in the figure includes an original image cylinder 3,
A motor 4 for rotating the original image cylinder 3 in the main scanning direction, a rotation angle encoder 5 for detecting the rotation angle of the original image cylinder 3, and a one-rotation encoder for detecting one rotation of the original image cylinder 3. 6, a scanning head 7 for scanning the color original image mounted on the original image cylinder 3, and a feed screw 8 for moving the scanning head 7 in the sub-scanning direction.
and a feed motor 9 for rotationally driving the feed screw 8.

On the other hand, the recording section 2 also has the same configuration as the original image scanning section 1, and includes a recording cylinder 10 and the recording cylinder 1.
A motor 11 for rotating 0 in the main scanning direction.
, a rotation angle encoder 12 for detecting the rotation angle of the recording cylinder 10, a one-turn encoder 13 for detecting one rotation of the recording cylinder 10, and a reproduction image on a film or the like attached to the recording cylinder 10. Recording head 1 for recording
4, a feed screw 15 for moving the recording head 14 in the sub-scanning direction, and a feed motor 16 for rotationally driving the feed screw 15.

The scanning head 7 is equipped with a color separation device 17, and the color separation device 17 converts the image signals obtained by scanning the color original into, for example, red (R), green (G),
It is output as a three-color separation signal of blue (B) and an unsharp signal (U). The three color separation signals and unsharp signal (U) from the color separation device 17 are
Next, the input is input to the color tone calculation circuit 18 used in a known color scanner, and after necessary processing such as logarithmic conversion, color correction gradation correction, detail emphasis, (magnification conversion), etc., the color tone of the color original image is converted into printed matter. Cyan (C) corresponding to the amount of ink of each color to express,
Magenta (M), Yellow (Y), Black (K) 4
Output as a color printing signal. On the other hand, the timing pulse generation circuit 26 generates a timing pulse (P ₁ ) and a one-rotation pulse (P 2 ) corresponding to the sampling pixels in the main scanning direction based on the pulse signals from the rotation angle encoder 5 and the one-rotation encoder ₆ . Each is occurring.

Now, when the M plate signal corresponding to the first scanning line is output from the color tone calculation circuit ₁₈ , the A/D
After being digitized by the converter 19, it is sequentially written into a buffer memory (FIFO) 22 via a switch 21.

On the other hand, C output from the color tone calculation circuit 18,
The Y and K version signals are digitized by the A/D converter 20 at twice the period of the M version signal and input to the subsequent multiplexer 24, but the M version signal is written to the buffer memory 22. While it is being loaded, 1
Since the multiplexer 24 is controlled not to operate by the rotation pulse _P2 , the C, Y, and K version signals corresponding to one scanning line are discarded.

Next, when the M-plate signal corresponding to the second scanning line is output from the color tone calculation circuit 18 via the A/D converter 19, the switch 21 is switched by the one-rotation pulse _P2 , so the M-plate signal is output. The signal is directly input to a compression circuit 23, which will be described later. At this time, the M version signal of the previous scanning line written in the buffer memory 22 is sent to the compression circuit 23 in the order in which it was written and in synchronization with the M version signal currently input to the compression circuit 23. 23, and the M version signals corresponding to a pair of pixels adjacent in the sub-scanning direction of both scanning lines are inputted to the compression circuit 23 as a pair. This compression circuit 23 performs encoding using non-linear quantization every time an M-version signal of, for example, 2×2 pixels corresponding to the compression unit area described above is prepared, and the compressed image signal m is sent to the multiplexer 23. is input.

On the other hand, since the C, Y, and K version signals are input to the multiplexer 24 at a cycle twice the scanning pixel pitch, the multiplexer 24 outputs the image signals grouped together for each compression unit area.
The data is output as a set of M ₀₀ , m' ₀₁ , m' ₁₀ , m' ₁₁ , C ₀₀ , Y ₀₀ , K ₀₀ , and sequentially written to an image memory 29 such as a magnetic disk via a buffer memory 25.

FIG. 6 is a block diagram showing an embodiment in which an image signal is compressed by the nonlinear quantization described above.

Now, buffer memory 22 and A/D converter 1
The M version signals input in parallel from 9 to the compression circuit 23 via the switch 21 are input to data selectors 40 ₁ and 40 ₂ , respectively, and from these data selectors 40 ₁ and 40 ₂ , the M version signals of each compression unit area are inputted in parallel to the compression circuit 23 via the switch 21. Signals M ₀₀ , M ₀₁ , M ₁₀ , and M ₁₁ are output in parallel. This M version signal is then sent to the adder 41 ₁ ,
41 ₂ and 41 ₃ , respectively, and prediction error signal values (M ₁₁ −M ₀₀ ), (M ₁₀ −M ₀₀ )(M ₀₁ −M ₀₀ ) are calculated. At this time, data selectors 40 ₁ , 4
0 ₂ are switched at a timing corresponding to the scanning pixel pitch, so each M version signal M ₀₀ , M ₀₁ ,
Although M ₁₀ and M ₁₁ are not all output at the same time, the inconvenience caused by this can be overcome by appropriately attaching a sample and hold circuit or a delay element to the illustrated data selectors 40 ₁ and 40 ₂ .

Prediction error signal values (M ₁₁ −M ₀₀ ), (M ₁₀ −M ₀₀ ), (M ₀₁ ) output from adders 41 ₁ , 41 ₂ , 41 ₃
-M ₀₀ ) are then stored in the table memories 42 _{1 and} 42 1 , respectively.
42 ₂ , 42 ₃ as address signals, and representative values m' ₁₁ , m' ₁₀ , m' ₀₁ as explained in FIG. 3 are output from each table memory 42 ₁ , 42 ₂ , 42 ₃ . .

As mentioned above, m' ₁₁ is expressed by 6 bits, m' ₀₁ and m' ₁₀ are each expressed by 5 bits, and these representative values m' ₁₁ , m' ₁₀ , m' ₀₁ and M ₀₀ (=8 bits) from sample hold circuit 43
is then input to the multiplexer 44, which outputs a compressed image signal m of a total of 24 bits (=3 bytes) at a timing corresponding to the pitch of the compression unit area.

The image signals of the plurality of original images compressed for each compression unit area as described above and written to the image memory 27 are then processed by an editing device (not shown) onto the image memory 27 or into another image memory. Based on the layout designation above, image data is rearranged according to the output position for each pattern.

The image memory 27 in which the edited image data has been written is set on the output side, and recording of a predetermined duplicate image is started.

That is, the image signal for each compression unit area read from the image memory 27' via the buffer memory 30 is converted into C, Y, K by the multiplexer 31.
It is divided into a plate signal and a compressed image signal,
The C, Y, and K version signals are directly sent to the decomposition version calculation circuit 33.
The compressed image signal m is input to a decomposition calculation circuit 33 via a reproduction device 32, which will be described later. In this separation calculation circuit 33, four color signals of CMYK are assigned to each pixel constituting the compression unit area, and at a timing corresponding to the scanning pixel pitch,
For example, four-color printing signals corresponding to two scanning lines are input to the buffer memory 34 in parallel. The buffer memory 34 is composed of at least two line memories (FIFO), and the four-color signal of the pixel corresponding to the scanning line to be recorded is output as is to the output control circuit 35, and the pixel corresponding to the next scanning line to be recorded is The four-color signals are written into the buffer memory 34 at any time. FIG. 7 shows an embodiment of the reproducing device 32 corresponding to the above-described FIG. 6, and will be briefly described below.

In FIG. 7, when an image signal m compressed to 24 bits (3 bytes) is input, the data selector 50 selects an _M00 signal corresponding to the representative value.
It is divided into m′ ₁₁ , m′ ₁₀ , m′ ₀₁ signals, and then
The signals m' ₁₁ , m' ₁₀ , and m' ₀₁ are input to table memories 51 ₁ , 51 ₂ , and 51 ₃ as address signals, respectively. These table memories 51 ₁ , 51
From ₂ and 51 ₃ , prediction error signals (M ₁₁ −M ₀₀ ), (M ₁₀ −M ₀₀ ), corresponding to m _{′ 11} , m′ 10 , and m′ ₀₁ respectively are obtained.
(M ₀₁ −M ₀₀ ) is output, and the next stage adder 52 ₁ ,
At 52 ₂ and 52 ₃ , they are added to the M ₀₀ signal, respectively, and signal values corresponding to M ₁₁ , M ₁₀ , M ₀₁ , and M ₀₀ are output.

(F) Effect The image signal compression method according to the present invention compresses the image signal responsible for the brightness, which was maintained in detail in all pixels of a color image in the conventional method, while ensuring image quality that can withstand practical use. However, since the signal amount responsible for brightness can be further reduced, the compression ratio of the image signal as a whole can be further increased.

Furthermore, in the image signal compression method according to the present invention, in each compression unit area, the brightness value of the representative pixel is not encoded, but only pixels other than the representative pixel are encoded. When performing image processing using the image signal, when displaying an image on a monitor with only specific pixels by thinning out some of the pixels of the original image, only the representative pixels can be used as is. This is advantageous because the monitor image can be immediately displayed on the monitor using the monitor, and there is no need to decode the encoded brightness values. In this way, according to the method of the present invention, a compressed image signal that is easy to handle can be obtained.

[Brief explanation of drawings]

FIG. 1 is a schematic diagram for explaining a conventional image signal compression method, and FIG. 2 is a schematic diagram for explaining an image signal compression method applying a predictive coding method according to an embodiment of the present invention. Figure 3 is a diagram showing an example of a transformation characteristic for taking a prediction error signal and applying nonlinear quantization, and Figure 4 is for explaining the case where an image is reproduced using the transformation characteristic shown in Figure 3. diagram,
FIG. 5 is a block diagram showing one example of implementing the compression method according to the present invention, FIG. 6 is an example of a compression circuit using nonlinear quantization, and FIG. 7 corresponds to FIG. 6. Examples of reproducing circuits are shown below. C ₀₀ , M ₀₀ , Y ₀₀ , K ₀₀ ...Each color plate signal at the representative pixel within the compression unit area, M ₀₁ , M ₁₀ , M ₁₁ ...
...The brightness value at pixels other than the representative pixel in the compression unit area when the M version signal is used as a signal responsible for brightness,
m′ ₀₁ , m′ ₁₀ , m′ ₁₁ ... Brightness value at each pixel position when the M version signal is compressed by predictive encoding.

Claims (1)

[Claims] 1 Set a compression unit area consisting of multiple pixels,
The representative pixel of the compression unit area is provided with all of the plurality of image signals corresponding to the color signals necessary when reproducing a color image, and each pixel other than the representative pixel of the compression unit area is provided with the plurality of image signals. In an image signal compression method that reduces the number of bits of the entire image signal obtained by scanning a color image by providing only the image signal corresponding to the brightness signal among the signals, a representative image of each compression unit area is used. The value of the image signal corresponding to the brightness signal to be given to a pixel other than the pixel is encoded by performing nonlinear quantization based on the difference between the value of the image signal and the image signal corresponding to the brightness signal of the representative pixel of the compression unit area. An image signal compression method characterized by: