JPH0553417B2 - - Google Patents

Description

[Detailed description of the invention]

[Industrial Application Field] The present invention relates to an adaptive quantizer used in a differential encoding device that digitizes a television signal and differentially encodes and transmits it. [Prior Art] Generally, in the quantizer of this type of differential encoding device, when the input signal of an image is n bits, the difference signal between the input signal and the predicted signal is n+1 bits or more. Therefore, the quantizer that quantizes and outputs the difference signal needs to convert at least the n+1 bit signal into a predetermined number of quantization levels. At this time, a large number of quantization levels were required in order to make quantization of flat areas and deterioration such as noise and overload less noticeable. Conventionally, as a method to improve this drawback, the input signal is amplitude-limited in advance and then differential encoding is performed, and a subtractor is used to obtain a difference signal from the input signal and the predicted signal, and a quantized output and the predicted signal are used. The operation of the adder to obtain the locally decoded signal from the signal is
The quantizer is n
There is a technique for quantizing a bit difference signal (for example, Bostelmann's folding quantizer, the contents of which are shown in German Patent Publication No. 2405534). [Problems to be Solved by the Invention] However, in the conventional Bostelmann folding quantizer described above, even if quantization noise due to quantization is added, the locally decoded signal (the input signal plus the quantization noise) In order to prevent the input signal (the same value) from exceeding the dynamic range of the input signal, the input signal is limited in amplitude to a range smaller than the dynamic range by the amplitude value of the maximum quantization noise, and then
It is necessary to perform differential encoding, and an amplitude-limited signal is output as the locally decoded signal and therefore the decoded signal on the receiving side. For this reason, when the amplitude of the maximum quantization noise is large, the range to which the amplitude is limited also becomes large, and as a result, the decoded signal has the disadvantage that the dynamic range is limited accordingly. SUMMARY OF THE INVENTION In view of the above drawbacks, it is a technical object of the present invention to provide a quantizer capable of obtaining a decoded signal having a dynamic range equivalent to that of an input signal not subject to amplitude limitations. [Means for Solving the Problems] According to the present invention, in a quantizer of a differential encoding device that quantizes and encodes a difference signal between an input signal and a predicted signal using a folding quantization method and transmits the encoded signal. , the upper limit of the dynamic range of the input signal, the range of values smaller than the upper limit by the maximum quantization noise of the quantizer, and the lower limit and the range of values larger by the maximum quantization noise from the lower limit. or means for outputting a converted signal by quantizing the input signal in accordance with a predetermined characteristic such that the area close to the lower limit is fine and the area far from the input signal is coarse, and the difference signal and the converted signal are , quantize the difference signal using the characteristics of the quantizer to obtain a quantized signal, and calculate the quantization noise added to the quantized signal by quantization and the level of the quantized input signal indicated by the converted signal. Determine whether the added value exceeds the dynamic range,
If there is a possibility that the dynamic range will be exceeded, the input signal is An adaptive quantizer characterized in that differential encoding can be performed using folded quantization without amplitude limitation is obtained. [Example] Next, the present invention will be described with reference to the drawings. FIG. 1 is a block diagram showing the configuration of one embodiment of the present invention. n bits input to input terminal 1,
For example, an input signal X with n=8 is supplied to the subtracter 2 and the conversion circuit 3. The quantization characteristic of quantizer 4 is symmetrical with positive and negative values, and the amplitude of the maximum quantization noise is ΔMX.
shall be. ^{In the} conversion ^circuit 3 ^, when ^the input signal ^-1 +ΔMX
If the value of X is within the lower limit range, a converted signal obtained by quantizing the value of On the other hand, in the case of an intermediate range that is not included in any of the ranges, the conversion circuit 3 outputs a conversion signal indicating that it is an intermediate range, and supplies it to the quantizer 4.
For example, if ΔMX is 32, input signal
The signal is quantized into signals with 16 different levels that are more than the level of , and a 4-bit signal is output as a converted signal. That is, if the upper limit is 127 and the lower limit is -128, then 127, 126, 125, 123, 119, 111, which are 0, 1, 2, 4, 8, 16, 24, and 32 away from this upper limit 103,
and -128, -127, -126, -124, -120, -112, −104,
−
Quantized to 96 levels. The intermediate range is quantized to a level 32 away from the upper and lower levels (127) (-128), ie, either 95 or -96 levels. In this way, the converted signal is the input signal
represents the level of quantization noise (ΔMX = 32) as long as the input signal
It can be seen that even when , the decoded signal does not exceed the dynamic range. Table 1 shows an example of the conversion characteristics of the conversion circuit 3.

【table】

[Table] The subtracter 2 uses the n-bit input signal X and the predictor 5.
It subtracts the n-bit prediction signal X^ supplied from the n-bit prediction signal X^ by an n-bit modulo operation, outputs an n-bit difference signal E, and supplies it to the quantizer 4. The quantizer 4 quantizes the n-bit difference signal E according to predetermined quantization characteristics and outputs a quantized signal Q representing the quantization level. However, quantization noise due to quantization Δ _q =Q
-E to the input signal (this value matches the local decoded signal Y=X^+Q) is likely to exceed the upper or lower limit of the dynamic range of the n-bit input signal A signal with a nearby quantization level, such as one higher or one lower, is output as a quantized signal Q, and the value X is obtained by adding quantization noise to the input signal X.
Quantization is performed adaptively so that +Δ _q (=X^+Q) does not exceed the upper or lower limit. Note that although the input signal X itself is not supplied to the quantizer 4, information on the quantized input signal X in the range necessary for determining whether to perform adaptive quantization can be obtained from the converted signal.
For example, when the conversion signal is 2 ^n-1 -1-4, the quantized output signal Q for the difference signal E is Q _i , and the quantization noise Δ _qi is 6, then X + Δ _qi exceeds the dynamic range. However, at the next lower quantization level Q _i-1 , the quantization noise Δ _qi-1 = Q _i-1 −E is usually negative, and X
+Δ _qi-1 does not exceed the dynamic range. Therefore, as the quantization output, a signal at the next lower quantization level is output as the quantization signal Q. To ensure that the value obtained by adding the quantization noise _Δq to the input signal X does not exceed the dynamic range of the input signal,
Quantized signal Q that adaptively quantizes the difference signal E
is output from the quantizer 4 and supplied to the code conversion circuit 7 and the adder 6. The code conversion circuit 7 converts the quantized signal Q into a code representing the quantization level, multiplexes it with information necessary for coded transmission such as a synchronization signal, converts it into a transmission line code, and outputs the code from the output terminal 8. The adder 6 adds the quantized signal Q and the prediction signal X^ to obtain a locally decoded signal Y (=X^+Q), which is supplied to the predictor 5. Through adaptive quantization, the quantized signal Q is selected so that the locally decoded signal Y does not exceed the n-bit dynamic range of the input signal X, so the adder 6 performs an n-bit modulo operation. Then, an n-bit locally decoded signal is output. The predictor 5 calculates and outputs an n-bit predicted signal X^ of the next sampling time according to a predetermined prediction characteristic from the locally decoded signal Y, and supplies it to the adder 6 and the subtracter 2. do. Next, a specific example of the quantizer 4 will be explained. The quantizer 4 has 4.5 bits (code conversion circuit 7
Now, encode the two samples together with 9 bits. ), the quantization steps are 0, 1, 2, 3, 4, 6, 8, 12, 16,
20, 24, maximum (minimum) quantization level is 96 (-96)
becomes. The quantization output Q for the difference signal E is shown below.
The quantization characteristics are shown in 2.

[Table] Adaptive quantization is performed as follows. When the converted signal for input signal X shows a positive value of 2 ^n-1 -1-Δ, the quantized signal Q for difference signal E is
Suppose that Q _i . If the quantization noise Δ _qi = Q _i −E is less than Δ, Q _i is output as is and Δ _qi
is larger than Δ, the one closest to Q _i is selected and output from among the quantized signals Q _ij such that Δ _qi-j =Q _ij −E is less than or equal to Δ. If the conversion signal has a negative value -
2 When ^n-1 +Δ is shown, if Δ _qi is greater than or equal to -Δ, Q _i is output as is, and if it is smaller than -Δ, Δ _qi-j is greater than or equal to -Δ. signal
Select Q _i+j and output. Q _-10 is used as the quantized signal Q that is one level higher than the maximum quantized output Q ₁₀ , and Q ₁₀ is used as the quantized signal Q that is one level lower than the minimum quantized output Q _-10 . Here, the relationship between Q ₁₀ and Q _-10 will be explained. In encoding using a folding quantizer, an adder for obtaining a locally decoded signal and a subtracter for obtaining a difference signal perform modulo operations using the same number of dynamic range bits as the input signal. A value exceeding the dynamic range is replaced by a signal of the same dynamic range as the input signal by a modulo operation. For example, if the input signal expressed in two's complement is 8 bits (signal value is -128 (10000000) to 127
(range of (01111111)), when the result of normal addition or subtraction is 128 (010000000) in 9-bit representation, when an 8-bit modulo operation is performed, the most significant bit (0) is ignored and the lower The value is determined to be -128 (10000000), which is indicated by only 8 bits. The difference signal input to the quantizer 4 is also obtained by modulo calculation, and when it is 128 in normal calculation,
The difference signal of the modulo operation is determined to be -128.
The quantizer 4 having the quantization characteristics shown in Table 2 outputs a quantized signal of Q _-10 (-96). In normal calculations, a difference signal with a value of 128 to 255 is input to the quantizer 4, and is determined to be a value in the range of -128 to -1, and a difference signal of -129 to -256 is determined to be a value in the range of -127 to 0.
It is determined that the value is within the range of . In other words, for differential signals exceeding the dynamic range, the quantization characteristic centered at 0 is shifted to the quantization characteristic centered at +256 or -256 shifted. . As a result, it can be seen that the quantized signal one above Q ₁₀ is Q _-10 , while the quantized signal one below Q _-10 is Q ₁₀ . For example, if ^the input signal
The converted signal is input to the quantizer 4. On the other hand, when E = 127 is input as the difference signal E, it is quantized to Q ₁₀ = 96 due to the quantization characteristics, and the quantization noise Δ _q =
Q ₁₀ −E=−31. In this case, the calculation of X + Δ _q uses the information of the conversion signal as _the value of X.
=-120-31, but since there is a possibility that X+ _Δq exceeds or exceeds the lower limit of the dynamic range, Q _-10 is selected as the quantization output Q _i as a quantization characteristic one higher than Q ₁₀ . At this time, the quantization noise Δ _q is
Q _-10 -E = 33 - 256, but considering that n = 8-bit modulo operation is performed when obtaining the local decoded signal, -256 in the second term is removed and Δ _q is equivalently +33. Become. Therefore, the value of X+Δ _q does not exceed the lower limit, and the quantizer 4 outputs Q ₋₁₀ as a quantized signal. At this time, since the quantization noise is at most 33, the locally decoded signal has a value close to the input signal X, and no large overload occurs. Next, the input signal X is -118 and the difference signal E is 83
The case will be explained below. Table 2 shows this difference signal E.
When quantized according to the quantization characteristics shown in , the quantized signal becomes Q ₉ (=72), and at this time,
The quantization noise (Δ _q =Q−E) is −9. The conversion signal of the input signal X has a conversion level of -120 from Table 1, and the conversion circuit 3 provides a conversion level of 1011 to the quantizer 4. The quantizer 4 converts this converted signal into an input signal X
is a signal with a value in the range of -120 to -113, and it is recognized that there is a level difference of 8 from the minimum conversion level of -120 to the lower limit of -128. When the quantizer 4 outputs a level value of 72 as a quantized signal, the quantization noise is -9, which is larger than the minimum level difference of 8 from the lower limit of the input signal There is a risk that the lower limit of is exceeded. In other words, the locally decoded signal contains -9 of quantization noise.
will be added, and since it takes a value in the range of -129 to -121, there is a possibility that the lower limit of -128 will be exceeded. In this way, if there is a possibility that the lower limit is exceeded, the quantizer 4 outputs the quantized signal _Q1 instead of 9.
It has the characteristic of outputting a higher quantized signal _Q10 . In this case, the quantization signal Q ₁₀ is 96 and the difference signal E is 83, so the quantization noise (Q
-E) becomes 96-83=12. The local decoded signal has quantization noise of 12 added to it, which changes from −120+12 to −113+12.
The signal has a value in the range of . Therefore, in any case, there is no possibility that the locally decoded signal exceeds the lower limit of -128, and the correct value will be reproduced. Furthermore, consider the case where the input signal X is 124 and the difference signal E is 83. In this case, the converted signal is 125 (0010) from Table 1, and the quantizer 4 recognizes that the input signal X is a signal having a value in the range of 124 to 125. On the other hand, when the quantizer 4 quantizes the difference signal E according to Table 2, the quantized signal becomes Q ₉ (72) and the quantization noise becomes -9. -9 of quantization noise (Q-E) is added to the locally decoded signal, and 124-9 to 125-
It will take a value in the range of 9, and there is no possibility of exceeding the upper limit of 127. Therefore, in this case, the value of 72 will be output as is as a quantized signal. Also, a case where the input signal X is 124 and the difference signal E is 62 will be explained. In this case, the converted signal is 0010 from Table 1, and in the quantizer 4, the input signal X is
It is determined that the signal has a value in the range of 124 to 125. When the difference signal E is quantized according to Table 2, the quantized signal becomes Q ₉ (72), and the quantization noise is
It becomes 10. Quantization noise of 10 is added to the locally decoded signal, resulting in a signal having a value in the range of 124+10 to 125+10, which exceeds the upper limit of 127. In this way, when the upper limit is exceeded or there is a possibility that it will be exceeded, the characteristics of the quantizer 4 are changed so that _Q8 , which is the next lower quantized signal, is output as the quantized signal instead of _Q9 . is determined. Therefore,
The quantized signal of quantizer 4 is Q ₈ (52), and the value is
If 52 quantized signals are output for 62 differential signals E, the quantization noise (Q-E) will be -10. A quantization noise of −10 is added to the locally decoded signal, resulting in 124
A signal having a value in the range -10 to 125-10 is output. In this case, there is no possibility of exceeding the upper limit of 127, and the correct value will be reproduced. Finally, the case where the input signal X is in the range from -96 to 95 will be explained. In this case, the converted signal is
Converts from 0111 to 1000. The maximum quantization noise (Q
-E) is in the range of ±32, so the locally decoded signal is in the range of -96±32 to 95±32, and the locally decoded signal is within the dynamic range (-128 to 127) due to the influence of quantization noise. Therefore, the quantizer 4 quantizes the difference signal E according to the quantization characteristics shown in Table 2, and outputs a quantized signal. If the converted signal is represented by 4 bits, n=8
The adaptive quantizer 4 for the bit difference signal is a ROM with a 12-bit address and an 8-bit output.
(Read Only Memory) is used to set the adaptive quantization characteristics for making the above-mentioned judgment in advance.
It can be configured by writing it into ROM and reading the value corresponding to the address. The current IC technology is a ROM with a 12-bit address, a 4-bit output, and an access speed of about 35nsee, and using this ROM can easily configure the quantizer 4. Note that the present invention is not limited to the embodiment shown here, and various configurations can be considered as long as the configuration is such that adaptive quantization is performed from the quantized input signal and the difference signal. In the above embodiment, only the adaptive quantizer was explained, but the code quantized by this quantizer can be decoded with the same configuration as the combination of adder 6 and predictor 5 in FIG. can be decoded by equipping the receiver with a receiver. More specifically, on the receiving side, first, the transmitted transmission path signal is separated to obtain information necessary for encoded transmission and a code representing the quantization level. The code representing the quantization level is sent to a code inverse conversion circuit, which performs code inverse conversion to obtain a quantized signal. The obtained quantized signal is sent to an adder in order to obtain a decoded signal with a quantized signal having the same value as the quantized signal output from the quantizer 4 on the transmitting side. The adder performs modulus addition of the quantized signal from the code inverse conversion circuit and the predicted signal from the predictor to obtain a decoded signal. As a result, this decoded signal will be the same as the locally decoded signal on the transmitting side. Therefore, the configuration of the receiving side will not be described in detail here. [Effects of the Invention] As explained above, the present invention adaptively performs quantization based on the relationship between the input signal and quantization noise, so that the input signal and Since the differential encoding process can be realized using modulo calculations with the same number of bits, the decoded signal is not subject to amplitude limitations, so it is possible to obtain the same dynamic range as the input signal. In addition, since the conversion circuit converts the input signal into a converted signal that extracts only the information necessary for the quantizer and then supplies it to the quantizer, the quantization is faster than when inputting the input signal as a converted signal. This has the effect of simplifying the configuration of the converter.

[Brief explanation of the drawing]

FIG. 1 is a block diagram showing the configuration of the present invention. 1... Input terminal, 2... Subtractor, 3... Conversion circuit, 4... Quantizer, 5... Predictor, 6... Adder, 7... Code conversion circuit, 8... Output terminal.

Claims (1)

[Claims] 1. In a quantizer of a differential encoding device that quantizes and encodes and transmits a differential signal between an input signal and a predicted signal using a folding quantization method, the upper limit of the dynamic range of the input signal and the above-mentioned For the range and lower limit of values that are smaller than the maximum quantization noise of the quantizer, and the range of values that are larger than the maximum quantization noise from the lower limit, areas close to the upper or lower limit are fine and areas further away are coarse. means for quantizing and converting the input signal according to predetermined characteristics to output a converted signal; and quantizing the difference signal from the difference signal and the conversion signal according to the characteristics of the quantizer. Find the quantized signal,
Determine whether or not the sum of quantization noise added to the quantized signal by quantization and the level of the quantized input signal indicated by the converted signal exceeds the dynamic range, and determine whether there is a risk of exceeding the dynamic range. In some cases, the input signal may be amplitude limited by means for selecting and outputting a signal at a nearby quantization level as a quantization signal instead of the quantization signal so as not to exceed the dynamic range. An adaptive quantizer characterized in that differential encoding can be performed using folded quantization even without the use of folded quantization.