JPS63158971A - Image compression system - Google Patents

Abstract

PURPOSE:To compress a color image and to shorten a transmission time, by applying quantization on the predicted remainder of a luminance signal and plural color difference signals. CONSTITUTION:In an image compressing and encoding circuit 10, the compressive encoding of the luminance signal and each of the color difference signals are performed. In such case, at a differential circuit 35, a first-order difference between the lines of the restored value of a preceding line or the predicted value of a present line from a prediction circuit 36 and the sample of the present line, is taken, and the first-order difference is encoded in a line direction by a variable sample density compression circuit 37. At this time, a second difference between the picture elements between a present sample and a preceding sample or the predicted value of the present sample can be obtained, and encoding is performed based on size relation with a quantization level. Therefore, since the directions of the first-order difference between the lines and the second- order difference between the picture elements are set in an orthogonal direction, correlation in a vertical and a horizontal directions can be eliminated.

Description

DETAILED DESCRIPTION OF THE INVENTION [Technical Field 1] The present invention relates to an image transmission method for transmitting image information in a narrow band.

[Background technology] Regarding color information, the International Commission on Illumination (CIE) color system XY
Z is,
．． In-edge, Z:λ=435. Although it is a single light color of 8n- blue, the three primary colors RGB of the NTSC system used in general television receivers have wavelength components distributed, and the peak wavelength is approximately R: 610 nm. G:
540 nm, B: 470n-. Below, RGB 3
The primary colors are the three primary colors of this NTSC system.

Now, in NTSC television broadcasting, the chromaticity difference visual acuity of a person's vision is less than 40% of the luminance visual acuity, and the chromaticity is only 19% of the luminance difference visual acuity for a color with a G-B difference.
Based on research by ord, brightness No. 46 (Y) is determined as Y=0.30R+0.59G+0.11B...(1)
and the color difference signal (Cn) is C+=(RY)/1.14...(2
)Cz=(B-Y)/2,03
...(3), the luminance signal is transmitted by AM modulation in the 4 MHz band, and the color difference signal is transmitted using the color subcarrier of 3.57 MHz.
9454MHz, quadrature two-phase modulation, approximately the luminance signal,
It is compressed and transmitted to 1/3 of the band.

In the receiver, the luminance signal (Y), color difference signal (CI), (C2
), the color difference signal C3 is demodulated into C,=(G-Y)X1.42-(4), and the three primary colors of RGB of the NTSC system are finally restored.

This is possible because the luminance signal (Y) includes the three primary colors RGB as shown in equation (1).

Color television cameras generally output color television combo knot signals that are modulated and synthesized from NTSC luminance signals, color difference signals, synchronization signals, and color burst signals for color synchronization of color subcarriers, and VTRs also use this camera. In addition, general color television cameras output the three primary colors of RGB and the same M signal, and
TSC method luminance signal (Y) and color difference signal (R-Y),
CG-Y) and outputs a synchronization signal. RG
Dedicated ICs for mutual conversion between B and color difference signals, modulation and demodulation using color subcarriers, and modulation and demodulation of color TV combo knot signals are commercially available for use in television cameras and image receivers.

In this way, in general television receivers, in order to show it to humans, the amount of information is reduced to 3J of RGB in the NTSC direction due to the characteristics of visual perception.
(It can be said that the color is compressed to about 1/2.

On the other hand, when transmitting color still images over a general analog telephone line, there is a method of sequentially FM modulating the three primary colors of RGB in the NTSC system, but there is a problem that it takes time to transmit because it is not compressed. For example, 256x256 pixels is 45
It is about seconds.

[Objective of the Invention 1] The present invention has been made in view of the above-mentioned problems, and its purpose is to provide an image compression method that can compress color images and transmit them in a short time. .

DISCLOSURE OF THE INVENTION J The present invention is an image compression method that compresses luminance information and color information of an image. It is characterized by compressing the residual using a variable sampling density or a quantization method such as DPCM.

Figure 1 shows the overall circuit configuration of an image transmission system using the method of the present invention.
The three primary colors of RGB in the SC system are divided into a luminance signal (Y) and color difference signals (R-Y) and (B-Y) by a color difference calculation circuit 2.

The A/D converter 3 converts the luminance signal (Y) into a digital value of 6 to 8 pixels)/ml by the color difference calculation circuit 2, and stores the converted digital value in the Y7 frame memory 4. .

This Y7 frame memory 4 has 512x512 pixels or 2
For example, 512X5 to store an image of 56X256m elements.
In the case of 12FW pixels, images of both odd and even fields are stored when an interlaced television camera is used, and in the case of 256×256 pixels, images of odd fields are stored.

Low-pass filters 6a and 6b each have a frequency of 500KHz
It constitutes a low-pass filter, which cuts high frequency components from the color difference signals (RY) and (B-Y) from the color difference calculation circuit 2. ), (B-Y) are A/D converter 7 respectively.
Dinotalized to 4 to 8 bits/pixel by m, 7b, 7-frame memory 8 of (RY), (B-Y)
rs of this frame memory 8m, 8b of this R-Y, BY stored in m, 8b! The brightness signal is 512X
In the case of 512 Kji elements, the number of pixels may be half each in the vertical and horizontal directions for the luminance signal, such as 256×256 pixels. However, it is possible to simplify the compression circuit by making it common for the luminance signal, or
In order to perform restored display using the /A converter, it is also possible to make the 7-frame memory for color difference have the same number of pixels as that for luminance.

Here, the timing circuit 13 creates a timing signal using the clock signal CLK from the color television camera 1 and the synchronization signal 5YNC, and based on this timing signal, the write trigger circuit 14 sends each A/D converter 3, 7a, 7b. Set the import timing of 7 dress counter 1.
5 to set the write address of each 7 frame memo '74.8m and 8b.

Now, in the embodiment shown in FIG. 11, the luminance signal (Y) and the color difference signal (R-
Y) and (B-Y) are selected line by line by the selector 9 and sequentially compressed and encoded by the image compression encoding circuit 10, and the modulation device l! 11, and is transmitted at a transmission rate of 9,600 bits/second through an ordinary telephone port a1.

Compressed codes can also be stored in computer storage devices, disk files, etc., reducing storage space. Call number 111 again! It is also possible to use the same computer data communication, modulation and decoding equipment when transmitting data using a computer, reduce response times for various control and monitoring systems, and improve reliability through error correction. But, FM
It has higher utility value than analog transmission that requires modulation.
The image quality is not inferior either.

Now, the compressed code transmitted via the telephone ml is transmitted to the demodulator f.
It is demodulated by i16 and further decoded by the image decoding and expansion circuit 1'1 to produce a luminance signal (Y), a color difference signal (R-Y),
(B - Y') is restored. The selector 18 receives the restored luminance signal (Y) and the color difference signal (RY).

(B-Y) and select Y frame memory 19, R-Y
7-frame memory 20a IB-Y are respectively stored in the addresses designated by the address counter 73 of the 7-frame memory 20b.

The digital value image data stored in these 7-frame memories 19.20m and 20b is transferred to the D/A converter 21.
, 22a, 22& convert into analog values, and the converted colors Mfff (R-Y), (B-Y) are smoothed by low-pass filters 23a, 23b, and then sent to a color difference modulation circuit 24 and a burst generator 33. It is modulated with the color subcarrier from the sync signal generator 32 and converted into a color television composite signal in the composite circuit 25 together with the sync signal from the sync signal generator 32, and can be displayed on a normal color monitor television 26 or
It is designed to be recorded in i'H. The wi signal from the synchronizing signal generator 32 enters the read trigger circuit 34, serves as a read trigger for the D/A converter 21*22m-22b, and also serves as a clock for the address counter 73.

By the way, the color restoration circuit 27 generates a color difference signal (G-Y) from the luminance signal (Y), color difference signals (R-Y), and (B-Y).
Y) and restore the three primary colors of RGB of the NTSC system.

Here, the color difference signal (G-Y) is G-Y=-0,51(RY)-0,19(B-Y)
...It can be expressed as (5).

When using a color television camera that outputs a general color television component No. 72148 as the color television camera/unit, the color difference calculation circuit 2 shown in FIG. 2 is replaced by the color television camera 1 as shown in FIG. A luminance signal (Y) is obtained from the color TV camera composite signal outputted from the amplifier 28, and a synchronization signal consisting of a vertical synchronization signal (SV) and a horizontal synchronization signal (SH) is separated by a synchronization separation circuit 31. Color difference signal (R-Y), (G-Y), Make (B-Y).

By the way, for the θ~4MHz band of the luminance signal (Y),
It is well known that due to the characteristics of visual perception, color difference signals are well defined in the band from 0 to 500 KH2.
) It is clear that the transmission band can be made narrower by transmitting the color difference signal in the same band as the luminance signal (Y). This is true for the R of the NTSC system.
When compressing and transmitting the three primary colors of GB, each of RGB requires a frame memory of the same N prime number as the luminance signal (Y), and the amount of code for each of the three primary colors of RGB is the same as that of the luminance signal (Y). On the other hand, in the case of a color difference signal, the number of pixels may be about 1/4 of the luminance signal (Y), which means that the luminance signal (Y) and two color difference signals (R-Y) are used.

The amount of information added to (B-Y) is about 1.5 times the luminance signal (Y), which is 1/1/2 of the three primary colors of RGB in the NTSC system.
It turns out that around 2 is good.

In the embodiment shown in FIG. 1, the color difference signals (R-Y) and (B-Y), which have fewer pixels than the luminance signal (Y), are D/A converted into analog values and then smoothly interpolated using low-pass filters 23m and 23b. RG of N'r SC method from luminance signal (Y)
The three primary colors of B will be restored.

Of course, the R-Y7 frame memory 20a%B-Y7 frame memory 20b is set to the same number of pixels as the luminance signal (Y), and the image decoding and expansion circuit 17 interpolates and restores the color difference (Ft (R-Y), (B-Y)). In this case, low-pass filters 23m and 23b
becomes invisible.

Figure 3 (a) shows the luminance signal (Y) and color difference signal (R-Y).
, (B-Y) is shown. Here, like the luminance signal (Y), the color difference signals (R-Y) and (B-Y) also have a correlation in the two-dimensional direction horizontal to the IR, so the second-order difference after removing this correlation is It has a frequency distribution centered on zero, and is more concentrated at zero than the first-order difference. Therefore, when this second-order difference is nonlinearly quantized using a power of 2, the quantization level close to zero becomes larger, improving image quality, and the compression rate becomes higher when using variable-length encoding or zero-code compression. Become.

Furthermore, with variable sample density encoding, as shown in Table 2, it is possible to increase the time difference value of a portion with a small quantization level, making it possible to lengthen the sample interval, thereby increasing the compression ratio.

Therefore, as shown in Fig. 3 (@), the image compression encoding circuit 1
0, the difference circuit 35 takes the line-to-line linear difference between the restored value of the previous line or the predicted value of the current line from the prediction circuit 36 and the sample of the current line, and uses this primary difference as a variable sample in the line direction. The density compression circuit 37 performs variable sampling density encoding. The variable sample density compression circuit 37 is a line buffer y37.
a, and a variable sample density encoding circuit 37b.

Now, the encoding in this variable sample density encoding circuit 37b is performed using the quantization characteristics shown in Table 2. Take the second order difference, and the value of this second order difference sign is the time difference value.
If it is smaller than the minimum value of the quantization level of 1", the sample interval is treated, and if the time difference value is 2", the quadratic difference between the two pixels is taken, which corresponds to the range of sample amplitude. It is encoded into a quantization level code.

Since the directions of the first-order difference between lines and the second-order difference between pixels are perpendicular, the correlation in the vertical and horizontal directions is removed.

In order to obtain the first-order difference between the lines of the next line, the value from the line buffer T41 is combined with the variable sample density decoding circuit 39m and the line buffer 3 according to the quantization level and time difference value.
The linear difference on the line is interpolated and restored from the decompressed and decoded value from the variable sample density expansion circuit 39 consisting of 9b, and the sample of the current line is restored in the interpolation synthesis circuit 40.

By removing redundancy in the code using the zero code compression circuit 38 of Table 4 or variable length coding, compression of the variable sampling density code becomes possible to some extent without degrading the image quality.

FIG. 3(b) shows the configuration of the image decoding and expansion circuit 17, in which the zero code expansion circuit 62 decodes a zero code or converts a variable length code into a fixed length variable sampling density code by variable sampling density decoding. The circuit 63 decodes it and loads it into the line buffer 64. Then, when the pixel is coarse, it is interpolated between lines in the interpolation synthesis circuit 65 using the line buffer 167 and the prediction circuit 66 using the compensation table shown in Table 5. Restore to line specimen.

Table 1 shows an example of quantization characteristics where the time difference value is around "1". In this case, variable sample density encoding is equivalent to DPCM,
The circuit of FIG. 3 is equivalent to the circuit of the tlSS diagram.

In the image compression encoding circuit 10 of this MS5 diagram (,), the register 42 is 1ii! The register 43, which is a register for 11 elements (or a memory element for delay), is a register for one line (or a memory element for delay), and is the same as the prediction circuit 36 in FIG. 3(a). That is, the first difference between the sample and the restored value of the front fin in the register 43 is obtained by the difference circuit 44,
A second-order difference signal is obtained between this first-order difference and one pixel of the restored previous line in a difference circuit 45, quantized in a quantization circuit 46, and compressed and encoded in a code compression circuit 47. Further, the second order difference restored by the expansion circuit 4 days is interpolated by an interpolation circuit 49 using one pixel from the register 42 and stored in each register 42 and 43. FIG. 5(b) shows the circuit configuration of the image decoding and expansion circuit 17. In this image decoding and expansion circuit 17, the third
Corresponding to the zero-code expansion circuit 62 shown in FIG. Restore.

Table 2 shows the quantization characteristics of the time rWI difference value "2", and in this case, the image compression encoding circuit 10 of FIG. 3(a) and the image compression encoding circuit 10 of FIG. 4(a)
It is equivalent to the circuit of N#) In the circuit of FIG. 4(,), the sample is 1 line, 2 pixels, and the restored value of the previous line from the fin buffer r50 is subtracted by the difference circuit 51 to become a 1 large difference Rana, and the register 53 stores the 1 pixel number. be done. The quantization circuit 52 compares the time difference value "1" with a pixel delayed by 11 pixels, and if the second-order difference from the difference circuit 54 is smaller than the minimum quantization level, it compares the second-order difference of the next pixel by one pixel. Outputs the sign of the sample amplitude. This code is encoded by the code compression circuit 5
It is compressed by 5. Further, this code is decoded by the expansion circuit 56, interpolated by the interpolation circuit 58 with the value of the register 57 for one pixel of the previous line, and stored in the register 60.

The register 57 associates the predicted value of the first-order difference with the register 53. The value of this register 57 is interpolated with the value of the register 58 for two pixels by an interpolation circuit 59 and stored in a line buffer 750, and a second difference is taken by a difference circuit 61 between it and the first difference. conversion circuit 52
Enter.

Table 3 shows an example of quantization characteristics with a time difference value of 4, in which case the fourth
Three registers 53 in the circuit shown in the figure (,) are connected in series,
If 57.633 registers are replaced, the circuit can be made equivalent to the circuit shown in FIG. 3(a). FIG. 4(b) shows an image decoding and decompression circuit 17 corresponding to the fourth el(a) circuit, and the fifth el(a) circuit is shown in FIG.
It is basically the same as Figure (b).

Table 4 also shows code examples for zero-kari code compression.

Now, when compressing color image data, for one line, the luminance signal (Y), the color difference signal (RY), and the color difference signal (B - Y) are compressed and encoded in the order of dispersion fins and are combined into one packet. encoded and transmitted as Prior to this compressed data, the camera number, image capture time, compression parameters, etc. are transmitted as a header packet. 1
The code of the packet is immediately decoded on the receiving side via the telephone line l, and compression, transmission, and decoding are performed on a packet-by-packet basis, so the sending side, transmission path, and receiving side receive and receive compression, transmission, and decoding in order. and transmits the next packet in parallel with the restoration.
Since the next packet is compressed and encoded in parallel with this transmission, the time from capturing the image to restoring it can be made approximately the same as the time required to transmit the image data.

The luminance signal (Y) and color difference signal (R
-Y) and (B-Y), so there is no need to transmit the color code.

Further, code information as to whether it is color or monochrome is transmitted in a header packet.

Figure 6 shows (RY), (B-Y) along with the luminance signal (Y).
, (G-Y), in which a frame memory 8c, an A/D converter 7c, and a low-pass filter 6c are used on the transmitting side, and a frame/module on the receiving side. j20c, D/A converter 2
2c and low-pass filter 6c, respectively, for color difference signals (G-Y
), it is placed in the extra sign.

In other words, the compression-encoded color difference signal (R-Y), (B-Y
) to restore the color difference signal (G − Y), 2
If the amplitude difference value is large due to nonlinear quantization of the order difference, a large error may occur. Therefore, in this example, in order to obtain the color difference signal (G - Y) more accurately, the color difference signal (G - Y) is encoded independently, and the Good color reproducibility of the system. Moreover, since the band of the color difference signal is narrower than that of the luminance signal (Y),
Since the band of the color difference signal is narrower than that of the luminance signal (Y), the amount of code is about 1/2 of that when transmitting 3g colors of RGB in the NTSC system, which has the effect of improving color reproducibility. In other words, the code amount of the color difference signal (G - Y) may be about 1/4 of the luminance signal (Y), the number of pixels may be 1/4 of the luminance signal (Y), and the RGB colors restored by the color restoration circuit 27 The signal is input to a color difference calculation circuit 2'' to extract the color difference signal and create a component knot signal.

(Table 1) (Table 2) (Table 3) (Table 4) (Table 5) Table 6) In the case of high-speed hardware with short calculation speed and memory access time for compression encoding, the three primary colors of the NTSC system are used. 7
By simply preparing three types of frame memories, a luminance signal (Y) and a color difference signal can be calculated and compressed and encoded in the same way, thereby realizing something equivalent to the present invention. Therefore, the configuration of the 7-frame memory is not limited to a configuration used only for color difference signals.

The number of pixels of the color difference signal is 1 compared to the number of pixels of the luminance signal (Y).
/4, the band of the color difference signal can be narrowed, and the amount of code can be reduced, but as shown in Tables 2 and 3, by setting the time difference value to 2 or more and using quantization characteristics with a high compression ratio, It is also possible to compress color difference signals. In other words, even if the number of pixels of the color difference signal is the same as the number of pixels of the luminance signal (Y), since the band of the color difference signal is narrow, the change will be gradual. This is because the compression ratio becomes higher.

In addition, the image transmission device can transmit coarse images in a short transmission time,
If you want the receiving side to raise the resolution of the screen displayed on the monitor TV and make it clearer by requesting sharpening as necessary, the sending side must remember the original image of the screen displayed on the receiving side. However, each color difference signal requires 1/4 the memory space of the luminance signal (Y), so only 172 memories are required when storing the three primary colors of RGB in the NTSC system.
This also applies if there are a plurality of television cameras, each with its own dedicated 7-frame memory, and the memory requirement can be reduced as well.

Next, the color difference signals (R-Y), (B-Y
)t(G-Y), each color difference signal has the following relationship with respect to the three primary colors of the NTSC system.

G-Y=-0,51(RY)-0,19(B-Y)
,...■Furthermore, C+ = (1/1.14)・
(RY), C2= (1/2,03)・(B-Y)
...■(t) (o, 3oo O, 590o
, ool(R1) On the other hand, NTSC television broadcasting uses a wideband color difference signal (■) close to human vision and a narrowband color difference signal (Q).These are color difference signals (R-Y),
The phase is advanced by 33 degrees with respect to (BY), and the relationship is expressed by the following equation.

1 =0.7+ (RY)-G, 2F (B-Y)
...■Q=0.48(RY)+0.
41 (B-Y) ・@Color difference signal band is 1
．． 5MHz1Q is 500KHz.

Therefore, in the configuration of the present invention, the R-Y7 frame memory 8a is
The frame memory is a B-Y7 frame memory 8beQ7 frame memory, and the bands of low-pass filters 6u and 6b connected in series to this frame memory are 1 and 5M, respectively.
Hz and 500 KHz, it corresponds to the color difference signal I%Q. Color modulation in analog processing is done by color difference signal engineering, Q
The color difference signal (R-Y) is also subjected to quadrature two-phase modulation (double balance modulation) like the color difference signals (R-Y) and (B-Y), so the color difference signal (R-Y)
modulator! When replaced with modulation, the color subcarrier (color burst) is modulated with a -57° phase, and when the B-Y modulator is replaced with Q modulation, the color difference signal modulation is modulated with a -90° phase.
The demodulation of the zero color difference compensation signals 1 and Q modulated by the color subcarrier of the transport is also performed with the color subcarrier having the same phase as the modulation, and the color difference signal (
The demodulator for R-Y) becomes a D demodulator, and the B-Y demodulator becomes a Q demodulator.

Restoration of the three primary colors according to the NTSC system is performed by a matrix circuit for the color difference signals (2) and (Q). That is, the color difference signal (R-Y
), (B-Y) is a color difference signal I,
It will be replaced with a "color restoration circuit" for Q, and the coefficients of the matrix will change.

It is said that color difference (N No. I, Q) is still deviated on the chromaticity diagram from the optimum value of the broadband axis and narrowband axis according to human vision.

In any case, the coefficients of the matrix change whether the color difference signal is defined by the 0 formula or 0 formula, or by a ring that is optimal for human vision, and in balanced modulation, the only difference is the phase and the weighting coefficient of the color difference signal. This applies to the gist of the present invention.

Effects of the Invention The present invention uses a luminance signal (Y) and a plurality of color difference signals to perform secondary prediction within a line of prediction residual between lines in an image compression method that compresses luminance information and color information of an image. Since the residual error is compressed using a variable purlin density or a quantization method such as DPCM, it is possible to narrow the band of the color difference signal compared to the luminance signal, reduce the number of pixels to about 1/4, and improve the quantization characteristics. The compression rate can be increased, which has the effect that the compression rate can be increased and the image quality can be transmitted without degrading the image quality compared to compressing and transmitting an RGB signal containing accuracy information while retaining the RGB three primary colors of the NTSC system. .

[Brief explanation of the drawing]

FIG. 1 is a circuit configuration diagram of Embodiment 1 of the present invention, FIG. 2 is a circuit configuration diagram showing an example of the circuit of the main part of the same, and FIG. 3(a) is a circuit configuration diagram of the image compression encoding circuit same as the above. , FIG. 3(b) is a circuit configuration diagram of the image decoding and decompression circuit same as above, FIG. 4(Jl) is a circuit configuration diagram of another example of the image compression encoding circuit same as above,
FIG. 5(b) is a circuit diagram of another example of the image decoding/expansion circuit same as above, FIG. 5(a) is a circuit diagram of another example of the image compression/encoding circuit same as above, and FIG. FIG. 6 is a circuit diagram of another example of the image decoding and decompression circuit same as above, and is a circuit diagram of a second embodiment of the present invention. 2... Color difference calculation circuit, 10... Image compression encoding circuit, 17... Image decoding/expansion circuit, 27... Color restoration circuit. Agent Patent Attorney Ishi 1) Figure 7, Figure 2 A;

Claims (4)

[Claims] (1) In an image compression method that compresses the luminance information and color information of an image, the luminance signal (Y) and multiple chrominance signals are used as a variable sample for the prediction residual between the lines and the second-order prediction residual within the line. An image compression method characterized by compression using density and quantization methods such as DPCM. (2) Color difference signals (R-Y) for the three primary colors of RGB
2. The image compression method according to claim 1, wherein all three primary colors of , (B-Y) and (G-Y) are compressed together with a luminance signal (Y). (3) The image compression method according to claim 1, wherein the color difference signal is compressed using pixels coarser than the luminance signal (Y). (4) The image compression method according to claim 1, wherein the color difference signal is compressed with respect to the luminance signal (Y) using a quantization characteristic with a high compression rate.