JPS639390A - Digital video transmission system - Google Patents

Abstract

(57) [Summary] This bulletin contains application data before electronic filing, so abstract data is not recorded.

Description

DETAILED DESCRIPTION OF THE INVENTION TECHNICAL FIELD The present invention relates generally to components for processing high quality video digital signals at transmitters and receivers of video transmission systems, and more particularly to apparatus for processing high quality video digital signals at the transmitters and receivers of such systems. The present invention relates to a differential pulse code modulation encoder and decoder for.

BACKGROUND OF THE INVENTION Broadcast or color video signals typically consist of three component signals: a luminance signal, called Y, which carries the contrast, or black and white, information of the image; Or (RY
) and (B-Y). A composite video signal, such as the NTSC video signal, which is the standard for color television broadcasting in North America,
These three analog component signals are fully combined for transmission. That is, a carrier wave is modulated in quadrature with both chrominance signals, and then the combined chrominance signal is interpolated with the baseband luminance signal. However, component video treats analog baseband and chrominance signals as separate channels and does not combine them in an all-analog format. Component video produces higher quality images compared to composite video. This is because component video avoids the crosstalk between these three analog components that is introduced when combining these components as a composite video, and component video is also wider for the chrominance component of the signal compared to the NTSC composite video format. This is to provide bandwidth.

Digital signal transmission is more resistant to noise and other image degradation introduced during transmission than analog signal transmission, and therefore the quality of the received video image is further improved by digital video signal transmission. The frequency, or bit rate, required for digital transmission of component video is determined by the sampling rate and number of bits per sample of the individual radiance f and chrominance signals. The sampling rate and number of bits used per sample are usually directly proportional to the resolution and quality of the image. In other words, the quality of the image is
It is usually proportional to the transmission bit rate and is therefore required to keep the transmission rate (frequency/frequency) at a maximum. but,
Conventional transmission equipment has a limit to the transmission speed that it can handle, and the cost of transmission equipment is usually proportional to the maximum transmission speed that it can handle. Therefore, it is required to keep the transmission speed to a minimum.

To resolve these conflicts, various techniques and devices have been proposed that are envisioned to reduce the transmission speed of component video signals. One example of how analog techniques can achieve this is by simply increasing the bandwidth of the chrominance signal (limiting it to, say, 1.5 MHz) in order to reduce the frequency at which the signal is sampled during digitization. (, which even limits the bandwidth of the luminance signal.

However, this method adds too much distortion to the video signal and
It also reduces resolution. Information about the details of the image is carried at high frequencies. Therefore, a lower signal bandwidth results in a loss of detail from the image.

Another example with analog technology focuses on the periodicity of the video signal's spectrum and samples the video signal at sub-Nyquist frequencies in order to reduce the overall sampling rate.

The sub-Nyquist frequency is twice the maximum frequency contained by the signal, and is theoretically the lowest frequency at which any signal can be sampled without losing all the information in the sampled signal. A disadvantage of the method is that sub-Nyquist sampling introduces distortion to the image due to aliasing, thus reducing the quality of the image.

In addition to these inherent problems, analog speed reduction methods have common disadvantages not found in digital speed reduction methods. Their main disadvantage is their poor signal-to-noise ratio, which in turn introduces noise and errors into the reproduced image. Another disadvantage common to analog speed reduction techniques is the difficulty in manipulating analog video signals. For example, sharp analog signal filters are not only difficult to implement, but also cause distortion to the video signal. Therefore, there is a growing trend to rely on digital techniques to reduce video transmission speed without significantly sacrificing the quality of the video image.

In the digital domain, that is, once an image is digitized, a variety of techniques can be used to compress the image. One of these techniques is interleaved subsampling. This method discards all samples of one phase of an image scanline (i.e., every other sample) and all samples of the opposite phase of an adjacent scanline, and instead of each discarded sample, Single K, vertically adjacent samples (
To interpolate the values of discarded samples at the receiver (should we use (() for horizontally adjacent samples (adjacent samples on the same scan line))? Another technique for interleaved subsampling is that the value of the chrominance signal is proportional to the value of the luminance signal. We only transmit the luminance interpolation code and use this code to interpolate both the discarded luminance and chrominance sample values.

Another compression technique is differential pulse code modulation (DPC).
M). This method reduces the bit deposit required to transmit the value of a sample by transmitting only information about the difference between the sample and its predicted value. This predicted value is derived from the value of the previous sample. This technique avoids the loss of image quality as a result of DPCM by using adaptive quantization. This adaptive quantization technique uses multiple "scales," or quantizers, to measure the difference between a sample value and its predicted value, as well as a It is used to select and use a "scale" that is expected to minimize the coarseness of the most important information to be layered. The selection of this "scale" is based on a value calculated from the predicted values of a plurality of previous samples. Implementation of this technology only partially achieves these objectives.

This is because the values of previous samples do not necessarily correctly predict the optimal "scale" to be applied to the current sample.

Although digital compression techniques are generally superior to analog techniques, they remove or only approximate some image information, thus reducing the quality of the image in this respect. Therefore, due to the need to maintain high image quality, to date it has not been possible to fully achieve the speed reduction provided by the combination of these techniques. In other words, transmission speeds for high quality images have remained unreasonably high. Therefore, what is needed in this field is to further reduce the transmission speed of high quality images without significantly reducing image quality.

SUMMARY OF THE INVENTION The present invention aims to overcome these and other difficulties of the prior art. According to one aspect of the invention, the digital encoding and decoding apparatus uses the value of the luminance signal to limit the value of the chrominance signal. An encoder and a decoder according to the invention for transmitting and receiving component video signals in digital form, respectively, convert the upper and lower limits of the values of the chrominance component video signal to the values of the corresponding luminance component video signal. an apparatus for predicting the value of a chrominance signal; an apparatus for determining whether the predicted value falls within the limits; and if the predicted value falls outside the limits. includes a device for employing all of the limit values closest to the predicted value as substitutes for the predicted value. In this way, a limited estimate of the chrominance signal is obtained.

The encoder further includes a device for generating a signal representative of the difference between this limited predicted value and the actual value of the chrominance signal. When the encoder is used in a transmitter device, a signal representative of this difference is transmitted over the transmission medium.

The thus transmitted signal traverses the medium and is received by a receiver device. The decoder of the receiver device includes a device for producing a recovered value of the chrominance signal by modifying this limited estimate by a difference value obtained from the signal representing this difference.

This encoder and decoder when used within a D P CM video transmission system! d double the color resolution at any transmission rate compared to the reproduced images obtained by conventional DPCM encoders and decoders. Therefore,
At any transmission rate, the quality of the transmitted image is greatly improved. On the contrary, traditional DPCM
Speeds that cannot be achieved by some encoders and decoders can be reduced. This means that the encoder and decoder can be used in conjunction with multiple transmission rate reduction techniques to increase total video transmission rates to rates unattainable to date while maintaining the image quality required for high-quality video. It is possible to drop it.

A high quality component video digital transmission system according to the invention combines this technique of limiting the chrominance signal with the luminance signal by other techniques, e.g. interleaving.
45MbP when used in conjunction with subsampling, chrominance prediction by luminance, and adaptive quantization.
s or less. This lower transmission rate is of great significance in that it allows high quality video transmission systems to use the conventional telephone T3 carrier as the transmission medium.

In another aspect of the invention, the adaptive unifying technique selects a plurality of "scales" for measuring the difference between a sample and its predicted value based on the predicted value of a previous sample. Let's do it based on the current sample only. The encoder according to the invention used in a device for transmitting a component video signal in digital form is characterized in that it provides a representative value of the quantization level of a first (i.e. one of luminance or chrominance) component video signal error. a plurality of sets, or "scales"; a device for selecting one set of error representative values based solely on the current predicted value of the first signal; 1, and includes means for making a selection based on the difference between the predicted and actual value of one signal, and a precautionary means for generating a signal representative of the error representative value thus selected. A transmitter device transmits this signal onto the medium.

The signals thus transmitted are received within a device for receiving component video signals in digital form.

A decoder used within this receiving device is configured to convert representative values of a plurality of sets of quantization levels of prediction errors of a first component video signal, one set of representative values to only the current predicted values of the first signal. an apparatus for selecting one error representative value from the selected set based on the signal representing the difference between all received signals; and apparatus for selecting a reproduced value of the first signal based on the predicted value. and an apparatus for generating by correcting with an error representative value.

When used within a DPCM video transmission system, the encoders and decoders described herein optimize the selection of a set of quantization level error representatives. Using these, a set representing the preferred quantization levels is selected. In other words, when reproducing a sample, it is possible to reproduce the entire appearance of the original sample, or to minimize the coarseness of the most important information (i.e.,
The level that maximizes resolution is selected. Therefore, the quality of the transmitted image is improved at any transmission rate. Conversely, the rate of transmission can be reduced to speeds that cannot be achieved by conventional adaptive quantization techniques without reducing the quality of the image below an acceptable level. Similar to the first aspect of the invention, this encoder and decoder can be used in conjunction with multiple transmission rate reduction techniques, thus:
It is now possible to reduce the total video transmission rate to rates unattainable to date while maintaining the image quality required for high quality video.

Preferably, the device for selecting a seratra of error representative values comprises a plurality of quantization groups, one of the predicted values of the first signal.
and an apparatus for selecting a set of representative values i based on the selected quantization groups. This method of quantizing predicted values completely eliminates the need to prepare a separate set of quantization level error representative values for each predicted value, thus reducing the coarseness of these directly representative levels. (This greatly reduces the number of representative value sets without increasing them. Also, according to this method, here too, the reproduction of the sample duplicates the entire appearance of the original sample [in order to minimize the coarseness of the most important information). In addition, different predicted values can be quantized into groups of different sizes.

Furthermore, preferably, the plurality of sets of representative values are comprised of a plurality of sets of representative values, each of which is associated with at least one predicted value.

Here, the error representative value of each set is determined by the maximum error value allowed by any predicted value of the at least one predicted value and the minimum error value allowed by any predicted value of the at least one predicted value. Represents a quantization level that subdivides the range of prediction error values of a signal.

By restricting the range of error values for each group of at least one predicted value in this way, the interval of the range for each individual group is limited to approximately half the total range of error values that includes the immediate error value of the possible predictions. Therefore, the coarseness of the quantization level represented by the individual set of Machler representative values is approximately halved, and the resolution of the transmitted reconstructed image is therefore approximately doubled.The resolution of the image is increased. The advantages of doing so are as explained above.

In an exemplary embodiment of a high quality component video digital transmission system, the DP at the transmitter
The CM encoder and the corresponding DPCM decoder at the receiver include respective adaptive quantizers. Here, the selection of a set of prediction error representative values each representative of all quantization levels of sample value prediction errors is performed based on the predicted values of only the sample currently being encoded or decoded. At each encoder and decoder, this predicted value is quantized into groups by a quantizer selector. A group is then selected based on the predicted values of the current sample only. An adaptive quantizer set is then selected based on the selected group. The selected set of predicted error representative values represents a quantization level that subdivides the range of error values bounded by the maximum and minimum possible error values of the selected group. Furthermore, the predicted value of each chrominance sample is limited by a limiter at each chrominance encoder and decoder to within the bounds determined by the corresponding chrominance sample value. Using all of these techniques together improves image quality, which allows transmission rates to be further reduced through the use of an interleaved subsampler that applies interleaved subsampling to both luminance and chrominance signal samples. shall be. At the transmitter, an interpolation selector selects an interpolation method for recovering sample values discarded during subsampling based only on the luminance signal samples. At the receiver, the selected method is used to recover both the luminance and corresponding chrominance values of the samples discarded by both the luminance sample interpolator and the chrominance sample interpolator.

Using these technologies together allows high quality component video transmission to be transmitted across the link at a full 45 Mbit/s.

These and other advantages and features of the invention will become more apparent when read in conjunction with the description and drawings of exemplary embodiments of the invention.

DESCRIPTION OF THE PREFERRED EMBODIMENT FIGS. 1 and 2 together show a system 2 for transmitting all component high quality video signals in digital form. FIG. 1 shows the transmitter portion of the system, while FIG. 2 shows the entire receiver portion. Although only one receiver section is shown, the system can include multiple receiver sections.

As shown in Figure 1, at the transmitter section the individual analog component video signals are converted from analog to digital form by sampling, subsampled by an interleaving scheme, and then KDPCM encoded. The subsampled luminance signal is further used to generate an interpolation code that is used to recover both the luminance signal and the corresponding chrominance signal that were discarded during subsampling. K is also used to limit the corresponding chrominance signal prediction value produced as part of the DPCM encoding of the chrominance subsamples.This chrominance DPCM code and the luma D P C
The M code as well as the interpolation code are then multiplexed and transmitted on a transmission link extending to the receiver section.

Analog component video signals are typically generated by horizontally scanning line by line across the image. These signals are Y at the transmitter part,
(RY), and (B-Y) are received in analog component format, each lead 101.

are applied to the input ports of analog-to-digital (A/D) converters 104, 105 and 106 through 102 and 103, respectively. A/D converters 104-106 are conventional devices; each performs band limiting of the video signal. a low-pass filter 160 for sampling the band-limited signal, a sampler 161 for sampling the band-limited signal, and a quantizer 162 for rounding the individual samples to the nearest digital level.The resulting 4:2
: To meet the two-component standard, the luminance signal (Y) is preferably band limited to a 4.2 MHz bandwidth, and the chrominance signals (R-Y) and (B-Y) are buffed to 2.1 MHz. and the band-limited luminance signal is sampled at a predetermined frequency that is preferably twice the sampling frequency of the chrominance signal. These band-limited signals are preferably sampled at super-Nyquist frequencies.

For example, the luminance signal is sampled at 10.35 MHz and the chrominance signal is sampled at 5.18 MHz. The output of the quantizer for each individual luma sample is preferably a conventional 8-bit unstretched pulse code modulated (PCM) digital Y, and the output of the quantizer for each individual chrominance sample is preferably a conventional 8-bit unstretched pulse code modulated (PCM) digital signal. The output is a PCM digital signal with a conventional 9-bit sign. Therefore, the individual outputs of A/D converters 104-106 are digital signal streams of the actual values of adjacent samples of the scan line of the image.

From A/D converters 104-106, the PCM signal streams are interleaved to subsamplers 107-10, respectively.
9 input. Each interleaved subsampler 107-109 is basically a switch with one input and two outputs. This switch channels these values of adjacent samples that it receives at its input to different ones of its two outputs. In other words, 1
The actual value of every sample of the first phase of one image scan line (i.e. every other sample) is switched to one output, and the actual value of every sample of the second phase of this scan line is switched to one output. is switched to the other output. The number of samples in each scan line is the same as the number of samples in the subsequent scan line, but with an opposite phase, so this switch changes the value of every sample in the second phase of the next image scan line. is switched to one output, and the values of all samples of the first phase of this scan line are switched to the other output (τ. Therefore, half of this sample value is switched to one output, and the value of all samples of the first phase of this scan line is switched to one output, The other half is switched to the other output, i.e. the value of one sample is switched to one output, and the value of this and the adjacent sample is switched to the other output.

This process is known as interleaved subsampling. These relationships are shown in FIG.

The values of the samples from one output of the interleaved sub-sampler 109 are transferred to the DP CM encoder 1 with a limiter.
12 is input. Interleave sub-sampler 1
The sample values from the other output of 09 are discarded. Therefore, only the values of every other adjacent sample are channeled to the encoder 112 with limiter.

The values of samples representing all the same points in the image as the samples whose values were discarded at sub-sampler 109 are also discarded at sub-sampler 108. Then, the sample value from the other output of the sub-sampler 108 is input to a DPCM encoder 111 having a limiter. Therefore, here too only the values of every other adjacent sample are channeled to the encoder 111 with limiter.

Because the luminance signal is sampled at twice the frequency of the chrominance signal, each chrominance sample has two corresponding luminance samples, ie, luminance samples that represent the same entire image area as the chrominance sample. The output of one of the subsamplers 107 ((the channeled luminance sample value is the value of a luminance sample whose sample value represents the same image point as all the chrominance samples that enter the subsamplers 108 and 1090; The luminance sample values channeled to the other output of subsampler 107 are the values of the luminance samples of the image points not represented by these chrominance samples.
The sample values from one output of the DPCM encoder 110
The sample value from the other output of the sub-sampler 107 is input to the interpolation selector 113.

That is, the chrominance samples encoded by DPCM and the luminance samples representing the same image point are
It is not PCM encoded and is represented by an interpolation selector encoder (5ELCODE), as will be explained in detail later. This relationship is also shown in FIG.

DPCM encoders 110-112 produce signals representing the actual value of each incoming sample by its difference from the predicted value. More specifically, for the actual value of each incoming sample, the DPCM encoder generates a sample predicted value from the value of the previous sample value and compares this predicted value with the received actual value; Generates code representing all the differences between the compared lines. The configuration of the luminance DPCM encoder 110 is shown in FIG. 4, while the configuration of the chrominance DPCM encoder along with limiters 111 and 112 is shown in FIG.

As shown in FIG. 4, luminance DPCM encoder 110 includes a conventional predictor 400. This predicts the value of a luminance sample from the reconstructed value of the previous luminance sample. Predictor 400 typically comprises a plurality of registers, each of which is used to store the recovered value of the previous luminance sample, and a word that combines the stored values to determine the current luminance sample Q.
It includes a logic circuit for generating a prediction Wk for.

The prediction III value outputted by the predictor 400 is sent to the difference circuit 4.
01, Quantizer selector (Q 5ELECTOR) 4
07 and is input to the addition circuit 403.

At the difference circuit 401, increment is performed from this predicted value.
The actual values of the degree samples received from the sub-sampler (see Figure 1) are subtracted. This result represents the difference between the actual value of this luminance sample and the predicted value, and thus represents any error in the predicted value. This error value is determined by the adaptive quantizer 402
is input.

The purpose of adaptive quantizer 4020 is to "round" this error signal to the nearest error representative value. Adaptive quantizer 4
The function of 02 is based on the relationship that exists between predicted sample values and actual sample values, as explained below.

The actual value of the luminance sample (/:l: represented by an 8-bit PCM digital signal. That is, the actual value of the sample is represented as any one of 256 values from 0 to 255. In the worst case, The exact opposite of the actual value.

That is, when the actual value is 255, the predicted r-direction will be 0 in the worst case, and if the actual value is O, the predicted f-direction will be 255 in the worst case. Therefore, the error value representing the difference between the actual value and the predicted value has a total of 511 possible direct ranges from -255 to +255. Assuming that this value is partitioned by, for example, 16 quantization levels, each with a unique error representative value specified by K, such as the level center value, the value of 511 will be reduced to only 16 error representative values. Rounding would be required and this quantization would be very coarse.O However, the present applicant is aware that any predicted sample value (C actual range of allowed error values + only half of the possible values of 4501, i.e. 256 values) We focused on the fact that the sum of the predicted value and the error value should be within the actual sample value, which ranges from 0 to 255 possible.
This is because the actual sample value should be within the range of 5. For example, if the actual sample value is 0, then the error value will only have values within the range 0 to 255; if the actual sample value is 127, the maximum error value allowed is +128; , the lowest error value allowed is -127, giving the full range from -127 to +128. Also, when the actual sample value is 255,
It has only values within the range of error value l1-255 to 0.

That is, the actual range of error values allowed for each predicted sample value is partitioned by all 16 quantization levels (thus, the width of each quantization level, and therefore the quantum represented by the typical error value at this level). The coarseness of the quantization is reduced by half of the above example.Therefore, it is necessary to use a full set of quantization levels different from each actual or predicted sample value.

However, Ganto Toyama further discovered that the quantization levels of this set could also be quantized, and that this trap could dramatically reduce the number of quantization levels without significantly increasing the level coarseness. Thus, for example, predicted sample values can be quantized into 16 groups, and all predicted values of a group can have an associated set of quantization levels. Since each group extends over a unique set of quantization levels, there are only 16 sets in total. The resulting quantization groups, quantization levels, and error representative values for these levels are shown in Table 1.

As indicated by the entries in Table 1, not all quantization levels have the same size. This occurs when the image is saturated around black or white, i.e. when the actual sample value is around 0 or 255, respectively.
This is based on the fact that the human eye has more difficulty discriminating between levels of contrast. That is, the quantization levels at which the reconstructed image values approach these saturation levels are given coarser spacing.

Based on the same eye characteristics, the predicted 11itH childization groups will also not all have the same size. In other words, quantization groups containing predicted values close to the saturation value are given coarser intervals.

As shown in Figure 4, the job of assigning a quantization group to a predicted value is the quantizer selector (Q5ELECT).
OR) 407. This Q 5ELE
The CTOR 407F is organized as a search table implemented as a read-only memory (ROM) 409. The ROM 490 selects one of the 256 ROM locations 493 for the 8-bit predicted value of the currently encoded sample.
is connected to address port 491. ROM location 4
93 is a 4-bit group code 464 that identifies the one of the 16 quantization groups to which that predicted value belongs.
including. The selected group code 464 is ROM4
90 to its data port 492 from where it is applied to adaptive quantizer 402 and level encoder 404.

Adaptive quantizer 402 is also organized as a lookup table and is
50. The ROM 450 stores a plurality of representative values 460 of signal prediction error quantization levels, 16 in this example. Here, each set corresponds to one of 16 quantization groups of predicted sample values.

Q5ELECTO input to adaptive quantizer 402
The four bits of the group code from R are connected to address port 451 of ROM 450 and constitute all of the most significant bits of the ROM address. This part of the ROM address selects one set 4600 of error representative values.

Ganto Toyama came to the conclusion that the current predicted value of the currently encoded sample is the preferred selector of the error representative value set. Therefore, the selection of the set is not based on previous predicted values, but only on the current predicted values. As mentioned above, this is based on the prediction of only the currently encoded samples.
This is achieved by selecting the predicted value singing group within the ECTOR 40,7.

The selected set 4600 representative values are selected by the error values input from the difference circuit 401 to the adaptive quantizer 102 in the ROM 450. This error-direct bit is R
Connected to address port 451 to OM450, RO
Constitutes the least significant bit of the M address. The addressed ROM location 453 stores the entire predetermined error representative value of the selected set 460 and the ROM 450 outputs this value at its data port 452.

The representative error value selected from the data port 452 of the ROM 450 is sent to the level encoder 404 and the adder circuit 403.
is input. Circuit 403 uses the error representative value to modify the predicted value.

The error representative value is added to the predicted value by a summation circuit 403 to generate a reconstructed value of the sample. The thus reproduced values are then input to the predictor 400. This also applies to the interpolation selector 113 and limiters 111 and 11.
It is also input to the 2-money DPCM encoder (see Figure 1).
.

The level encoder 404 converts this error representative value into a DC, PM code that identifies all the representative values. Encoder 404 is also organized as a lookup table and stores multiple sets 48 of DPCM codes.
This is implemented as a ROM 470 that stores 0. In this example, there are 16 sets of 16 error representative values, so ROM 470 stores all 16 sets 480 of 16 DPCM codes. Each DPCM code is a 4-bit code in this example since 4 bits are required to represent 16 different error representative values.

These four bits of group code from QSELECTOR input to encoder 404 are connected to the address port of ROM 470 and constitute all of the most significant bits of the ROM address. This portion of the ROM address selects one of the 16 sets 4800 of DPCM codes. (As explained later, there is no need to send this information to the receiver section since the receiver section generates multiple copies of this part of the ROM address.) This bit of the error representative value is also
It is also connected to the address portion of 470 and constitutes the least significant bit of the ROM address. This address is ROM470
Predetermined DPCM code within? All storage locations 473 are selected, and the ROM 470 outputs this D P CM code to its data port 472 .

FIG. 5 shows a chrominance DPCM encoder with a limiter. Chrominance encoders 111 and 112 have the same structure and function as luminance encoder 110, except for some enhancements. The basic operation of encoders 111 and 112 is based on the relationships that exist between the predicted and actual values of the chrominance samples. These relationships are essentially the same as those described for luminance. However, the actual The actual and predicted values of chrominance can be negative, extending from -155 to +255. Therefore, the possible range of chrominance error values is -510 to +255. +510, which has 1021 different values.However, as in the case of luminance, the sum of the predicted value and the error representative value is within the possible range of the actual value for any particular predicted value. The possible range of chrominance error values for is half of the total possible range, or 512 different values.

Since this range is twice the range of the luminance error range, this means that for any number of error range quantization levels, the quantization level of the chrominance error value is twice coarser than the quantization level of the luminance error value. It looks like it does.

However, Ganjin Toyama believes that (R-Y) and (B-Y) chrominance have unique characteristics that can be used to eliminate this difference in roughness at the quantization level, and therefore provide better results. It has been discovered that chrominance reproduction and/or reduction of transmission bit rate can be realized. This property is (Y), (R-
It arises from the relationship between the Y) and (B-Y) signals,
This relationship is tabulated below.

Y=, 3R+, 59C;+, 1IBR-Y=
7R-, 59G-, 11BB-Y= 3R-, 59
G+, 89B From this relationship, I discovered that the dynamic range of the (R-Y) and (B-Y) signals is bounded by the value of the luminance signal. The characteristics of the range of the (RY) signal are plotted against the range of the Y signal (C) in Figure 6, while the characteristics of the range of the (B-Y) signal are plotted against the range of the Y signal in Figure 7. These figures show that for any value of luminance, the range of chrominance values is limited to a maximum of 256 possible values.
Limits the entire range of dynamic values, which halves the chrominance dynamic range. As a result, the chrominance for any particular value of chrominance
The range of possible error values is halved. That is, maximum 2
There are 56 different values. Therefore, if the values of the luminance samples are used to limit the range of values of the corresponding chrominance samples according to the characteristic ζC of FIGS. 6 and 7, then the coarseness of the chrominance error quantization level is is equal to the roughness of

Returning to Figure 5, the chrominance encoder is chrominance
Encoder 11 for generating predicted values for samples (C
0 includes a predictor 500 similar to predictor 400 of 0. Since the predicted value is based on the values of other previous samples, it falls somewhere within the full dynamic range of the chrominance signal. Therefore, this predicted value is input to the limiter 50'5. Limiter 505 limits the predicted value to the range determined by FIGS. 6 and 7, and derives from this a luminance-limited predicted chrominance direct. The limiter 505 includes both an upper limit detection circuit 502 and a lower limit detection circuit 521. Circuits 520 and 521 realize the characteristics shown in FIG. 6 for encoder 111 and the characteristics shown in FIG. 7 for encoder 112. This property is implemented as a lookup table in ROM. Circuits 520 and 521 are ROMs 530 and 543, respectively.
including. This ROM includes locations 533 and 543, respectively. This position stores the upper limit value and lower limit value, respectively. ROMs 530 and 50 store the upper and lower limit values of the chrominance samples for the possible variations of the corresponding luminance samples, respectively. Luminance DPCM encoder 11
The reproduced values of the luminance samples obtained from 0 <ROM 530 and 5 to select positions 533 and 543, respectively.
40 addresses are input to ports 531 and 541. Locations 533 and 543 store nine corresponding chrominance upper and lower limits, respectively, and these values are R
OM530 and 531 data ports 532 and 54
It is output to 2.

From circuits 520 and 521, this limit value is input to comparators 522 and 5230, respectively, and multiplexer 524.
added to the input.

The predicted value is applied to other inputs of comparators 522 and 523 and to a third input of multiplexer 524. Comparator 52
The output of comparator 523 indicates whether the predicted value exceeds the upper limit, and the output of comparator 523 indicates whether the lower limit exceeds the predicted value. The outputs of comparison layers 522 and 523 are the inputs of selection logic 525 (
can be added with An input of logic 525, in turn, is connected to a select (SEL) input of multiplexer 524. The comparator 522 indicates that the predetermined TfI11 value is the upper limit e! Logic 525 causes multiplexer 524 to select and output the upper direct tone. If comparator 523 indicates that the lower limit exceeds the predicted value, logic 525 causes multiplexer 524 to select and output the lower limit direct tone. In other words, if the predicted @ is outside this limit, the limiter 505 adopts all the limit values closest to the predicted value instead of the earthwork 1]@.

Then, when the predicted value is within the limits, logic 525 causes multiplexer 524 to select and output the predicted a++ value. The output of multiplexer 524, and thus of limiter 505, is therefore a uniform predicted chrominance value.

From this limited predicted value, the actual value is subtracted in difference circuit 501 to obtain the error value. This error value is input to adaptive quantizer 502.

To the limited predicted value of the chrominance output by the limiter 505, a luminance reproduction value is added in an adder circuit 506 to obtain a corrected chrominance value. The purpose of this addition is to 6 and 7) and ensure that the value falls within the range O to 255. This corrected value is then input to the KQ S ELECTOR.

Q5ELECTOR 507 is the Q of the luminance encoder 410
Similar to 5ELECTOR407. This outputs a full 4-bit group code identifying the quantization group to which the input correction value belongs. This group code is then input to an adaptive quantizer 502 and a level encoder 504.

Quantizer 502 extends on a similar structure to adaptive quantizer 402 of encoder 510. The chrominance signal is compared to the luminance signal ψ
DPCM encoder 111
The ROM of the quantizer 502 and 112 has the same contents as the ROM 450 of the quantizer 402 of the DPCM encoder 110.

However, the human eye is not sensitive to fluctuations in the chrominance signal or to fluctuations in the luminance signal. Based on this fact, it is also possible to double the coarseness of the chrominance quantization level and thus halve the number of levels and the number of error representative values that all this level specifies, in which case the adaptive quantizer 50
2 is included in 16 sets of 8 error representative values. 8
In this case, the DPCM code output by encoder 504 is simply a 3-bit code, since only 3 bits are required to represent the value .

The group code from the Q5ELFJCTOR 507 constitutes the most significant bit of the ROM address of the adaptive quantizer 502 and selects one of the set of error representative values therein. (As in the case of brightness, the receiver part generates a copy of this part of the ROM address and therefore similarly (there is no need to transmit this information to the whole receiver part). Difference circuit 5
The error values from 01 to 01 form the lowest order of ROM addresses and are used in the adaptive quantizer 502 to select one of the selected set of error representative values. The error representative value selected in this manner is outputted by the adaptive quantizer 502 to the encoder 504 and the adder circuit 503.

In addition circuit 503, the error representative value thus selected is added to the limited predicted value to obtain the chrominance reconstruction value. The reconstructed direct is then input to the predictor 500.

Encoder 504\i Encoder 4 of luminance DPCM encoder 110
Similar to 04. This converts the error representative value to a D P 0M code. In this example, since there are 145 representative values in each set of chrominance error representative values, encoder 504 converts the error representative values into 4-bit codes. Q5ELE input to encoder 504
Group code from CTOR507 is set DPC
Select all M codes, and select a code that identifies all M codes from this selected set of codes as the error representative value.

This selected DPCM code is then output by encoder 504.

Returning to FIG. 1, the interpolation selector 113 takes as its input every other actual sample of luminance (direct and reproduced sample value of luminance) and provides an interpolation code (SELCODE) k for the inputted actual sample value. Interval selector 1 to output
13 is a well-known electrical manufacturer. Receive brightness Sanfurno 1'
For the actual values of JI, the interpolation selector 113 interpolates all the reproduced values of adjacent samples horizontally, vertically, diagonally left, and diagonally right. These interludes are illustrated in FIG. 3 and are designated as H, V, LD and RD, respectively. The interpolation selector 113 then performs a full comparison of the interpolated value with the actual value to determine which interpolation method gives the closest value to the actual value. The interpolation selector 113 then generates a 2-bit interpolation code (SEL) that fully identifies the interpolation method.
CODE) Full output and discard the actual value. Information carried by a 2-bit interpolation code? Another way to interpret it is that this information identifies two samples whose values are interpolated to reproduce the value of one adjacent sample.

DPCM encoder 110-112 and interpolation selector 113
The code output by multiplexer 114K
input and combined into a single digital signal stream. A multiplexer 114'/'i transfers this signal stream to a buffer 115.
Add to. Transmitter 116 retrieves this signal from buffer 115 and transmits it over transmission link 200 with a transmission rate of 45M)(z).Buffer 11
5 fully smooths the transmission rate, even if the rate of the output of multiplexer 114 is bursty, e.g., when only samples representing the visible portion of the video source are encoded (i.e., ``active-only'' video). Its function is to ensure that the signal stream is output on link 200 at a constant rate.

Returning to FIG. 2, this signal stream transmitted on link 200 is decomposed into component signals at the receiver. The DPCM encoded signal is decoded into reproduced values of subsampled luminance and chrominance signals. This recovered value of the subsampled luminance signal is used to limit the predicted value of the chrominance signal generated as part of the DP CM encoding process. The interpolation code (SELCODE) is used to obtain the values of the subsampled luminance and chrominance signals discarded at the transmitter section by interpolation along with the reproduced values of the subsampled luminance and chrominance signals. The interpolated and regenerated signals are then combined into regenerated luminance and chrominance signals and converted from digital to analog format.

The signal stream transmitted on link 200 is received by receiver 216. Receiver 216 applies this signal stream to buffer 215. This signal is retrieved from here by demultiplexer 214. Buffer 215 allows demultiplexer 214 to retrieve this signal at a different rate than the steady transmission rate on link 200, eg, in burst mode. Demultiplexer 214 separates this signal stream into its component codes. This brightness DPC
The M code is input to a DPCM encoder 210, and the chrominance DPCM code is input to limiters 211 and 21, respectively.
2 inputs into one DPCM encoder.

Decoders 210-212 are entered to recover sample values from the DPCM code. More specifically, a DPCM decoder converts an incoming DPCM code into an error representative value identified by the code, derives a sample's predicted value from the value of a previous sample value, and converts the predicted value by the error representative value. Obtain the reproduction value of the sample by correcting it. Brightness D
The configuration of the PCM decoder 210 is shown in the eighth step, while:
Chrominance DPCM with limiters 211 and 212
The configuration of the decoder is shown in FIG.

As shown in FIG. 8(/'C), the luminance DPCM decoder 2
10 includes a conventional predictor 800. The predictor 800 predicts the compensation of the luminance sample from the value of the previous luminance reproduction sample. Prediction 6so.

fully replicates the predictor 400 of the encoder 110 to verify that the predicted value for sample 1 (is always the same at both the encoder 100 and the decoder 210).

Prej y++= Prejl11 output by s o o
(Direct is quantization selector (Q 5ELECTOR) 80
7 and adder circuit 803VC are input. QSELECT
OR807 is a duplicate of Q5ELECTOR407. This group code output is the demultiplexer 21
The adaptive quantizer 80 as well as the luminance DPCM code from 4
2 is input.

Quantizer 803 is the inverse of encoder 404.

This returns the DPCM code to the error representative value represented by that code. The quantizer 802 has a table search ROM 87 containing 16 sets 880 of 16 error representative values.
This is realized as 0. The group code output by Q5ELECTOR807 is 1 of set 880.
while the DPCM code selects one of the error representative values of the selected set 880. The error representative value output by quantizer 802 in response to the DPCM code is the same as the error representative value output by quantizer 402 and resulting in the DPCM code generated within DPCM encoder 110.

At the addition circuit 803, this error representative value is sent to the predictor 80.
0 is added to the predicted value to generate the reproduction for sample (C).

This recovered value is input to circuitry outside of predictor 800 and decoder 210. As explained later (, decoder 21
The luminance reproduction value generated by 0 is the basis for reconstructing the values of other luminance and chrominance samples by interpolation.

As shown in FIG. 9, a chrominance DPCM decoder with limiter includes a conventional predictor 900'. Predictor 900 performs a direct prediction of a chrominance sample from the value of a previous chrominance sample. Predictor 900 replicates predictor 500 of the encoder of FIG. The predicted value output by predictor 900 is input to limiter 905, which is a replica of limiter 505 of FIG. Limiter 905 takes as a second input the luminance reproduction value from decoder 210 and produces a limited predicted value.

This is input to adder circuits 906 and 903.

The function of adder circuit 906 is the same as that of adder circuit 506 in FIG. Circuit 906 adds the reproduced luminance value from decoder 210 to the limited predicted value and generates the resulting correction value e.
Output to Q5ELECTOR907.

Q 5ELECTOR907F QSELEC in Figure 5
Copy TOR507. The group code thus generated is sent to the adaptive quantizer 9 as well as the DPCM encoded chrominance error code from the multiplexer 214.
Added to 02.

Quantizer 902 is similar to quantizer 802 and encoder 50
It is the opposite of 4. This converts the DPCM error code into an error representative value. The group code output of QSELECTOR 907 is held by quantizer 902.
Select one of the 16 sets of error representative values of the selected set, while the DPCM code is 1-:)'? of the error representative values of the selected set. select.

The error representative value output by quantizer 902 in response to the DPCM code is the same as the error representative value output by quantizer 502 and resulting in the generation of the DPCM code within the DPCM encoder of FIG.

In adder circuit 903, this error representative value is added to the predicted value to generate a reproduced value for the sample.

This reconstructed hypothetical value is applied to predictor 900 and circuitry outside the decoder of FIG.

Returning to FIG. 2, the reproduced values output by decoders 210-212 are transmitted to multiplexers 204-206, respectively.
, and are input to interpolators 207-209, respectively.

Interpolators 207-209 are conventional.

This is done at the subsampler and interpolation selector of FIG. Get the value of the sample with the discarded value. Although in FIG. 1 the interpolation code was calculated only for luminance, it is used in FIG. 2 to obtain the direct tones of both the luminance and chrominance samples. Luminance-based predictions of chrominance typically involve spatial variations in luminance as well as spatial variations in rominance, and therefore
This is because sample values tend to have the same relationship as adjacent luminance samples that correspond to each other. Thus, each of the interpolators 207-209 performs a full interpolation, and uses the adjacent values specified by the interpolation code to reproduce the reproduced values of adjacent samples obtained from the decoders 210-212. interpolate sample values.

Since the reproduction value of every other sample is obtained by interpolation, the value obtained by interpolation (C is
are input into multiplexers 204-206, respectively, where they are combined with the reproduced values obtained by decoders 210-212K, respectively. Multiprefusa 20
4-206 generates a stream of values, where the values of adjacent samples alternate with the individual values generated by decoders 210-212 and the direct tones generated by interpolators 207-209, respectively. arranged in the correct order.

The output streams of multiplexers 204-206 still have digital PCM encoding format. Therefore, these are each D
/A converters 201-203 reconvert to analog form.

D/A converters 201-203 are conventional devices. Each digital-to-analog domain converter, e.g., R-2R network 190, followed by a first
Low-pass filter 19 of the corresponding A/D converter in the figure
Is it the same passband as 1? Low bass filter with 19
Contains 1. The output of the D/A converter is Y', (RY)'
, and (B-Y)' are reproduced analog luminance and analog chrominance signals. These are the signals Y, (RY), and (B
-Y).

It will be apparent to those skilled in the art that various changes and modifications can be made to the exemplary embodiments described.

These changes and modifications may be made without departing from the spirit and scope of the invention or diminishing any of the advantages obtained thereby. It is therefore intended that such changes and modifications be covered by the appended claims.

[Brief explanation of drawings]

FIG. 1 is a block diagram of a transmitter portion of a digital component video transmission system as an example fully embodying an example of the present invention; FIG. 2 is a diagram showing a receiver portion of this transmission system; FIG. A diagram showing the operation of the sampler, sub-sampler and interpolation selector in full space: Figure 4 is a block diagram of the luminance DPCM encoder of Figure 1; Figure 5 is a block diagram of the chrominance DPCM encoder with the limiter of Figure 1;
A block diagram of the M encoder; Figure 6 shows the relationship between the luminance (Y) and chrominance (RY) sample values; Figure 7 shows the relationship between the luminance (Y) and chrominance (CB-Y) sample values; Figure 8 is a block diagram of the rational DPCM decoder of Figure 2; and Figure 9 is a block diagram of the chrominance DPCM decoder of Figure 2.
A complete block diagram of the CM decoder is shown. [Explanation of symbols of main parts] A/D conversion 6 104.105.106 Interleaved sub-sampler 107.108. 1.
09 Luminance DPCM encoder 110.111.11
2MUX 114DMUX
214 Luminance DPCM Decoder 210. 211.
212 Interpolator 207.208.20
9D/A converter 201.202.203 Applicant: American Telephone & Telegraph Company Procedural Amendment August 7, 1985 Commissioner of the Patent Office Kuro 1) Mr. Akio 1, Indication of Case 1985 Patent Application No. 151854 Item 2. Name of the invention Digital video transmission system 3. Relationship with the case of the person making the amendment Patent applicant 4. Agent 5. Subject of the amendment ``Description'' 6. Contents of the amendment As shown in the attached sheet 1, b+, +t ㅅj Please submit one copy of the certified letter A, which has been engraved as shown in the attached sheet.

Claims (1)

[Claims] 1. An apparatus for transmitting a component video signal in a digital format, wherein the apparatus includes, for each chrominance video component signal: a first method for predicting the value of the chrominance video component signal; a second computing device for determining upper and lower limits of the value of the chrominance signal based on the value of the corresponding luminance component video signal; a second computing device coupled to the first and second devices; a third arithmetic device for determining whether the predicted value of the chrominance signal falls between the upper and lower limits; a fourth arithmetic unit for obtaining a limited predicted value of the chrominance signal by adopting the limit value closest to the predicted value as a substitute for the predicted value of the chrominance signal when the predicted value of the chrominance signal deviates from between; and a fifth computing device coupled to the fourth device for generating a signal representative of the difference between the limited predicted value and the actual value of the chrominance signal; a sixth computing device for generating a signal representing a value of the luminance video component signal; and transmitting the generated signal coupled to each of the fifth device and the sixth device; Apparatus characterized in that it comprises a seventh apparatus for transmitting on a medium. 2. The device according to claim 1, wherein each of the fifth devices: a representative value of a plurality of sets of quantization levels of chrominance prediction errors; an eighth apparatus for selecting a set of said plurality of chrominance error representative values coupled to said apparatus based solely on said limited predicted values of said chrominance signal; said selected set and said fifth apparatus; Selecting a chrominance error representative value from the selected set of chrominance error representative values coupled to the second device based on the difference between the limited predicted value of the chrominance signal and the actual value of the chrominance signal. and a device for generating a signal representative of the selected chrominance error representative value in response to the selected value. 3. The apparatus of claim 2, wherein each of said eighth apparatuses: a plurality of quantization groups of predicted values of a chrominance signal; coupled to said plurality of groups and a fourth apparatus; an eleventh arithmetic unit for selecting one quantization group based solely on the limited predicted values of the chrominance signal; and in response to the selected group coupled to the plurality of sets; Apparatus comprising a twelfth apparatus for selecting a set of chrominance error representative values based on the selected quantization group. 4. The apparatus of claim 2, wherein each of the plurality of sets of error representative values: comprises a plurality of sets of chrominance error representative values, each of which is associated with at least one predicted value of the chrominance signal. , the error representative value of each set is the maximum error allowed by any predicted value of the at least one predicted value and the minimum error allowed by any predicted value of the at least one predicted value. Apparatus characterized in that the apparatus represents a quantization level that subdivides a range of chrominance prediction error values defined by the values. 5. By the apparatus of claim 2, each of the eighth apparatuses: a set of said chrominance error representative values coupled to said plurality of sets and said fourth apparatus Apparatus characterized in that it comprises an eleventh apparatus for selecting based solely on the sum of the limited predicted value of the chrominance signal and the value of the corresponding luminance signal. 6. The apparatus of claim 6, wherein the apparatus further: is coupled to the first apparatus and responsive to the selected value, for at least one video component signal; a sixth error device for recovering a first sample value from a predicted first sample value of the signal and the selected signal error representative value; the signal coupled to the sixth device; a seventh arithmetic device for interpolating the recovered values of at least one first sample of the signal to obtain a plurality of interpolated values of second samples of the signal; an eighth operation for comparing which of the plurality of interpolated values of the second sample of the signal coupled to is closest to the actual value of the second sample of the signal; a device for channeling actual values of adjacent samples of the signal coupled to the eighth device and the third device to the third different device and the eighth device; An apparatus characterized in that it includes a ninth arithmetic unit. 7. An apparatus for receiving a digital component video signal, wherein for each chrominance video component signal: a first computing unit for predicting the value of the chrominance video component signal; a corresponding luminance component; a limit on the value of the chrominance signal based on the value of the video signal; determining whether the predicted value of the chrominance signal coupled to the first and second devices falls between the upper and lower limits; a third arithmetic unit for the chrominance signal; when the predicted value coupled to the third unit deviates from between the upper and lower limits, a calculation unit closest to the predicted value is substituted for the predicted value of the chrominance signal; a fourth computing device for obtaining a limited predicted value of the chrominance signal by employing the limit; and for generating a reproduction value of the chrominance component video coupled to the fourth device; a fifth computing device for modifying the limited predicted value of the chrominance signal by an error value represented by the received first signal; the device further comprising: a luminance video component; a sixth computing device for generating a reproduction value of a signal based on the received second signal; and a transmission medium coupled to the fifth and sixth devices; a seventh computing device for receiving a second signal and transmitting the received first and second signals to the fifth and sixth devices, respectively. 8. The apparatus of claim 7, wherein each of the fifth apparatus comprises: a plurality of sets of quantum-level representative values of chrominance prediction errors; the plurality of sets and the seventh apparatus; a ninth apparatus for selecting one chrominance error representative value from the selected set of chrominance error representative values coupled to the received first signal representing the chrominance error representative value; and chrominance component video signal by modifying the limited predicted value of the chrominance signal by the selected chrominance error representative value in response to the selected value coupled to the fourth device. A device characterized in that it includes a tenth device for generating a reproduction value of. 9. The apparatus of claim 8, wherein each of the eighth devices: a plurality of quantization groups of predicted values of the chrominance signal; coupled to the plurality of groups and the fourth device; an eleventh computing unit for selecting one quantization group based only on the limited predicted values of the chrominance signal; and in response to the selected group combined with the plurality of sets; A twelfth computing device for selecting a set of chrominance error representative values based on the selected quantization group. 10. The apparatus of claim 8, wherein each of the plurality of sets of error representative values: comprises a plurality of sets of chrominance error representative values, each of which is associated with at least one predicted value of the chrominance signal. , the error representative value of each set is the maximum error allowed by any predicted value of the at least one predicted value and the minimum error allowed by any predicted value of the at least one predicted value. Apparatus characterized in that the apparatus represents a quantization level that subdivides a range of chrominance prediction error values bounded by the values. 11. By the device according to claim 8,
Each of the fourth devices may: simply combine the set of chrominance error representative values coupled to the plurality of sets and the fourth device with the limited predicted value of the chrominance signal and the corresponding value of the luminance signal. and an eleventh device for selecting based on the sum of . 12. In the device according to claim 11,
The apparatus further comprises: for at least one video component signal: the recovered value of at least one first sample of the signal specified by a received code coupled to the fourth apparatus; a sixth arithmetic device for obtaining a reproduced value of a second sample of the signal adjacent to the first sample of the signal by interpolation; and the fourth and sixth device. Apparatus characterized in that it comprises a seventh arithmetic unit for producing a stream of reconstructed values of every other first and second sample of said signal coupled to said signal.