JPH02253784A - Encoder - Google Patents

Abstract

PURPOSE:To widely improve a processing speed without lowering forecasting accuracy by parallelly providing plural encoder circuits to generate a forecasting value by using picture elements, which are arranged in a direction to cross the horizontal direction of a picture, and respectively processing sampling values corresponding to the picture elements to be arranged in the horizontal direction in the plural encoder circuits as mentioned above. CONSTITUTION:A sampling value Xi inputted to a picture input terminal 30 is impressed on D-type flip-flops 32 and 34, divided into an even number-th picture element X2i and an odd number-th picture element X2i+1 and respectively impressed to DPCM encoders 36 and 38. Then, encoding is parallelly executed. For a Y2i to be outputted from the DPCM encoder 36 and a Y2i+1 to be outputted from the DPCM encoder 38, the both signals are made serial by being suitably selected by a switch 40 and an output signal Yi is outputted to an output terminal 42. Thus, the speed of processing to be executed by the DPCM encoders 36 and 38 is made double and a double high band signal can be processed without lowering the forecasting accuracy.

Description

DETAILED DESCRIPTION OF THE INVENTION [Field of Industrial Application] The present invention relates to an encoding device that encodes a difference value between a sample value and its predicted value.

[Prior Art] When image information, audio information, etc. is digitized and transmitted, encoding is performed to reduce the amount of information to be transmitted.

One of the encoding methods is differential encoding (Differential encoding).
There is a method using alPulse Code Modulation (hereinafter referred to as DPCM), which compresses the amount of information by utilizing the correlation between adjacent sample values.

Specifically, the coded sample value is decoded, the decoded value is used to obtain the predicted value for the next subsampled sample value, and the difference between this predicted value and the input sample value is quantized. A method of encoding is used.

FIG. fi5 is a block diagram showing a configuration example of a general encoding device based on the above principle. Further, FIG. 3 shows an explanatory diagram showing a sampling pattern by the sub-sampler 3 of FIG. 5.

1 is an input terminal to which input data Xi (for example, a digital television signal) is applied. 2 is input data X
, a prefilter that attenuates the two-dimensional spatial frequency domain in
3 is a subsampler that applies offset subsampling to the output signal of the prefilter 2, and 4 is a DPCM encoder that encodes the output signal of the subsampler 3.

5 is a transmission path, 6 is a DPCM decoder that decodes the encoded signal from the DPCM encoder 4 and performs band expansion processing, and 7 is an interpolation filter that performs interpolation on the signal expanded by the DPCM decoder. , 8 is the output data■.

This is an output terminal that outputs .

In the above configuration, for example, an 8-bit digital television signal is applied to the input terminal 1, and a pre-filter 2 attenuates a predetermined two-dimensional spatial frequency region.
Offset subsampling is performed by the subsampler 3. This offset subsampling
Processed as shown in the pattern shown.

In FIG. 3, the solid lines represent the scanning lines of the first field, and the broken lines represent the scanning lines of the second field. Here, it is assumed that the input signal is a 2:1 interlaced television signal. This television signal is sampled in a grid pattern at points indicated by O and × marks, and the sub-sampler 3 transmits only the sample points (pixels) marked by O (in addition, fl is the sampling frequency). In this way, we perform the well-known offset subsampling in a staggered manner,
By interpolating the sample points that were not transmitted by the subsequent interpolation filter on the receiving side, the image can be transmitted without reducing the horizontal and vertical resolution.

However, the resolution in the diagonal direction deteriorates, but human visual characteristics are such that the resolution in the diagonal direction is smaller than that in the horizontal and vertical directions.
Prefilter 2 does not cause much deterioration.
The characteristics of are designed to attenuate the spatial frequency in the diagonal direction, and prevent the aliasing distortion caused by the above-mentioned subsamples.

The signal subsampled by subsampler 3 is D
For example, a signal of 8 bits per sample point (pixel) is compressed into 4 bits by the PCM encoder 4 and sent to the transmission path 5. The 4-bit encoded signal from transmission path 5 is D
The signal is decoded by a PCMPCM decoder and expanded to the original 8-bit signal. Next, the interpolation filter interpolates the sample points indicated by x marks in FIG. 3 that were not transmitted to the transmission path 5 based on the surrounding sample points that were transmitted. The interpolated output data V is output from the output terminal 8.

Next, details of the configuration of the DPCM encoder 4 shown in FIG. 5 will be explained. FIG. 4 is a block diagram showing an example of the configuration of a general PCM encoder 4. As shown in FIG. In this encoding, the encoded sample values as described above are once decoded (local decoding), and the decoded values are used to obtain predicted values for the next sample value.

A method is used in which the error between the predicted value and the actual value is quantized and encoded.

10 is an input terminal into which the sample value Xi is input; 12 is a subtracter that performs subtraction between the sample value This is a quantizer that sends out the DPCM code Yi to the output terminal 16. 1B is an inverse quantizer that combines the DPCM code Y into a difference value; 20 is an adder that adds the predicted value P and the output signal of the inverse quantizer 18 to restore the difference value to a sample value; A limiter 1 and 24 limits the amplitude of the output signal of the adder 20 to a predetermined range and obtains the local decoded value V, which is output from the limiter 22 to the next sample value X, etc. This is a D-type friub flop that applies the predicted value Pl+1 to the subtracter 12.

Next, the operation of the DPCM encoder 4 having the configuration shown in FIG. 4 will be explained.

A subtracter 12 subtracts the predicted value PI from the sample value X input to the input terminal IO. The difference value due to this subtraction is
Quantized by the quantizer 14, the DPCM code Y,
It is output to the output terminal 16 as Further, the output of the quantizer 14 is applied to the inverse quantizer 18, where the DPCM code Y is composited with the difference value, and then output to the adder 20.

The adder 20 adds the predicted value P output from the D-type flip-flop 24 to the difference value. As a result, the difference value (f
fi child-ized representative value) is restored to a sample value (locally decoded value). The restored sample value is applied to a D-type flip-flop 24 after its amplitude is limited to a predetermined range by a limiter 22 .

The D-type flip-flop 24 outputs the locally decoded value outputted from the limiter 22 in synchronization with the clock, and outputs the predicted value P1.1 for the next sampled value X1.1. ．． is applied to the subtracter 12 as .

Generally, the distribution of difference values of predicted values is biased toward small values, and compressed transmission of information becomes possible by encoding and transmitting the difference values using nonlinear quantization.

[Problems to be Solved by the Invention] In the conventional encoding device as described above, the previous value, that is, the value two sampling points before, is used as the predicted value, so when subsampling is performed, the difference between the sampling points As the spatial distance increases, the prediction accuracy decreases. For this reason, distortion due to quantization appears in the image, resulting in edge business (deterioration of image quality at the edges) and granular noise (roughness of the reproduced image in flat areas).
2, making it difficult to transmit high-quality images.

A method to solve this problem is to use not only sample values on the same horizontal scan line but also sample values on other scan lines to generate predicted values, so-called two-dimensional prediction, and improve prediction accuracy. I can think of it.

However, in the above-mentioned encoding device that performs the two-dimensional t' side, it is difficult to perform parallel processing that encodes a plurality of pixels simultaneously without reducing prediction accuracy.

For example, if you divide all pixels into multiple groups for each pixel or each scanning line and try to encode each group in parallel, the pixels to be encoded and the pixels used to generate predicted values are separated by the screen J. As a result, the prediction accuracy decreases significantly.

Therefore, an encoding device that performs this type of two-dimensional prediction is difficult to handle high-speed processing, and is not suitable for processing digital video signals with extremely high bit rates.

The present invention was made to solve these problems, and aims to provide an encoding device that improves prediction accuracy, enables high-speed processing, and is capable of encoding wideband video signals. It is something to do.

[Means for Solving the Problems] In order to achieve the above object, the present invention provides an encoder that encodes a difference value between an input sample value obtained by sampling a video signal and its predicted value, and the encoder A plurality of encoding circuits each consisting of a predictor that generates the predicted value using sample values aligned in a direction intersecting the horizontal direction on the screen for pixels to be encoded are provided in parallel, and The plurality of encoding circuits respectively process sample values corresponding to a plurality of pixels arranged in a row.

[Operation] According to the encoding device configured as described above, pixels to be used for generating prediction A1 can be arbitrarily selected, and a predicted value is generated using the sample value of the pixel closest to the pixel corresponding to the input sample value. Can be generated. That is, since the distance between the input sample value and the predicted value on the image can be shortened, the prediction accuracy can be improved. Further, by processing a plurality of pixels arranged in the horizontal direction in separate encoding circuits, encoding can be performed in parallel. Therefore, it is possible to encode high-band image signals at a speed that is several times faster than the processing speed of each encoder.

[Example] As mentioned above, in the conventional encoding apparatus shown in FIG. The spatial distance between sample points increases and prediction accuracy decreases. For this reason,
Distortion due to quantization appears in images, leading to deterioration such as edge business (deterioration of image quality at the edges) and granular noise (roughness of the reproduced image in flat areas), making it difficult to transmit high-quality images.

Therefore, in order to solve this problem, the encoding circuit shown in FIG. 2 is used in the apparatus of the embodiment of the present invention.

The encoding circuit shown in FIG. 2 uses an IH (i-water period) delay device 26 in place of the D-type flip-flop 24 of the DPCM encoder 4 shown in FIG.

Incidentally, in FIG. 2, the same reference numerals are used for the same components as in FIG. 4, and therefore, redundant explanation will be omitted below. Furthermore, the process of generating the DPCM code Y and decoding it using this is the same as in the case of FIG. 4, so the explanation will be omitted.

Next, the operation of the encoding circuit having the configuration shown in FIG. 2 will be explained.

The local decoded value V output from the limiter 22 is input to the IH delay m26, 1111! The t= line 26 delays the locally decoded value Vi by l horizontal scanning line portion, and outputs it to the subtracter 12 as the predicted value P. This predicted value Pi
is not the previous value, but a sample point one horizontal scanning line above on the screen.

In the encoding apparatus shown in FIGS. 4 and 5, the distance on the image between the sample value X1 and the predicted value Pl (hereinafter referred to as the predicted distance) takes the previous value and is 2/f.

However, in this configuration, it becomes 2/ν. In this case, the sampling frequency f1 is set to be slightly higher than twice the transmission bandwidth of the television signal.

The range of the input (difference value) of the quantizer 14 is the predicted value P1
If we know 255~P, then op, (however, the sample value
, is protonated with 8 bits). Therefore, depending on the predicted value P, the characteristics of the quantizer 14 and the inverse quantizer 18 can be optimized within the range at that time, and a sliding quantizer is constructed.

This makes it possible to reduce the quantization step to 1/2, making it possible to perform high-quality encoding with less quantization noise.

However, if the encoding circuit configured as shown in FIG. A large capacity ROM (4 bytes) for the encoder 18 (
4 Normally, such ROMs have slow access times, so 1 For example, when trying to encode high-bandwidth signals such as high-definition,
Due to insufficient processing time, it was not possible to encode high-band signals as it was.

FIG. 1 is a block diagram showing the main configuration of the apparatus of this embodiment.

In FIG. 1, 30 is an image input terminal to which a sample value X is applied, 32 is a D terminal input to the sample value X, and φ
A D-type flip-flop 34 receives the sample value Xi as a clock signal and inputs the sample value Xi as a D terminal input. This is a D-type flip-flop that uses the clock signal as the clock signal. 36 and 38 are DPCM encoding circuits having the configuration shown in FIG.
The output signal of the D-type flip-flop 32 is applied to the CM encoding circuit 36, and the output signal of the D-type flip-flop 34 is applied to the DPCM encoding circuit 38. 40 is D
Output of PCM encoding circuit 36 and DPCM encoding circuit 38
42 is an output terminal.

Next, the operation of the embodiment with the above configuration will be explained.

The sample value X input to the image input terminal 30 is.

Applied to D-type flip-flops 32 and 34. The D-type flip-flop 32 receives a clock signal φ synchronized with the input timing of even-numbered pixels. triggered by D
The type flip-flop 34 receives a clock signal φ synchronized with the input timing of odd-numbered pixels. triggered by As a result, the sample value X is divided into an even-numbered pixel X2+ and an odd-numbered pixel X2i,+1.

The divided sample values X are applied to the DPCM encoders 36 and 3B, respectively, and encoding is performed in parallel. DP
Since the CM encoders 36 and 38 perform prediction only in the vertical direction, pixels in the horizontal direction are processed independently of each other. Y2. output from the DPCM encoder 36.
and Y 2 i output from the DPCM encoder 38
+l is appropriately selected by the switch 40, so that both signals are serialized, and the output signal Y or the output terminal 4
2 is output.

With the above configuration, the DPCM encoders 36 and 3
8 doubles the processing speed, as shown in FIG. 5. Compared to the case using two DPCM encoders, it is possible to process twice as many high-band signals without reducing prediction accuracy.

Incidentally, in the above embodiment, an example was shown in which two encoders are parallelized, but the present invention is not limited to this, and any number of encoders can be parallelized. As the number of parallel processes increases, it becomes possible to increase the processing speed.

In the above embodiment, the sample points of each field are arranged in a grid pattern by field offset subsampling of the video signal, so pixels adjacent in the vertical direction of the screen are used to generate predicted values. When a signal is subjected to line offset subsampling, the closest pixels on the screen are arranged diagonally. In this case, prediction accuracy can be improved by using sample values corresponding to pixels diagonally above the pixel to be encoded to generate a predicted value. Further, in this case as well, processing speed can be increased by processing a plurality of pixels arranged in the horizontal direction using a plurality of encoding circuits provided in parallel.

[Effects of the Invention] As is clear from the above, according to the present invention, when encoding a difference value between an input sample value and its predicted value, the predicted values are aligned in the j direction or intersecting the horizontal direction of the screen. A plurality of encoding circuits are provided in parallel to generate data using the pixels arranged in the horizontal direction, and each of the plurality of encoding circuits processes sample values corresponding to the plurality of pixels arranged in the horizontal direction. This allows the encoding circuit to perform parallel processing, making it possible to significantly increase processing speed without reducing prediction accuracy.

[Brief explanation of drawings]

FIG. 1 is a block diagram showing the main configuration of an embodiment of the encoding device of the present invention, MS2IA is a block diagram showing the configuration of the DPCM encoding circuit in FIG. 1, and FIG. 3 is a sub-sampler in FIG. 5. FIG. 4 is a block diagram showing the configuration of the DPCM encoder shown in FIG. 51M, and FIG. 5 is a block diagram showing an example of the configuration of a conventional encoding device. In the figure. 3: Subsampler 6: DPCM decoder 12 subtractor 14/quantizer 18: inverse quantizer 20: adder 22 collimator 26: 1) 1 delay line 32.34. : D-type flip-flop 36.38: D P CM encoding circuit 40: Switch P, : Predicted value ■, : Local local signal value: Sample value X2. : Even numbered pixel X 2f*1. : Odd numbered pixel Y, : DPCM coat Figure 1 Agent Patent attorney ■1 Harusaku Kitatake 2 Figure

Claims (1)

[Claims] An encoder that encodes the difference value between the input sample value obtained by sampling the video signal and its predicted value; A plurality of encoding circuits each including a predictor that generates the predicted value using the sample values of the aligned pixels are provided in parallel, and the sample values corresponding to the plurality of pixels aligned in the horizontal direction are generated by the plurality of encoding circuits. An encoding device characterized by processing each.