JPH1056638A - Image coder, image decoder and image coding/decoding device - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an image coder in which nonlinear quantization processing is conducted at a high speed with a small circuit scale without use of a ROM table. SOLUTION: A prediction value generating section 106 generates a prediction value from picture elements around an input picture element, a linear quantization generating section 102 and a nonlinear quantization generating section 103 set a quantization representative value based on the predicted value, and a quantization section 104 selects and outputs a quantization representative value. Thus, nonlinear quantization is executed without the use of a quantization table and the input picture element is directly quantized, then the effect in a transmission line error is very decreased.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

The present invention relates to an image encoding device, an image decoding device, and an image encoding / decoding device.
In particular, an image encoding device that compresses a TV signal by image encoding and records it on a recording medium such as a VTR, an image decoding device that reproduces compressed data recorded on the recording medium, and an image encoding device that performs recording and reproduction The present invention relates to a decoding device.

[0002]

2. Description of the Related Art When obtaining digital data that can be handled by a computer or the like based on a video signal such as a TV signal, generally, first, image data obtained by digitizing a video signal is obtained, and data obtained by compression-encoding the image data is recorded or transmitted. Or to be done. The digitized image data is obtained as a sequence of pixel data having pixel values representing luminance and color difference, and is encoded by arithmetic processing to obtain encoded image data.

[0003] As for compression encoding of image data based on a video signal, there is prediction encoding as a compression method with high reproduction image quality. The predictive coding is a method of generating a predicted value for an input pixel to be coded, and transmitting data obtained by nonlinearly quantizing a difference value between the input pixel and the predicted value. Regarding the predicted value, when dealing with an image from a video signal, the pixel value of a certain point is closer than the neighboring pixel based on the fact that pixel values representing luminance and color difference are likely to be the same or close to each other. Is obtained by predicting Such predictive coding has been widely used because the circuit scale of the device is small and the compression rate is low, that is, when the data rate after compression is high, high image quality can be obtained.

FIG. 29 is a conceptual diagram for explaining linear and non-linear processing. The input data has a certain dynamic range. That is, it can be represented within a dynamic range of d bits and can be linearly processed. When trying to obtain n-bit output data for this input data by quantization, an appropriate number of quantized representative values are selected, and a quantized value is assigned to the representative value. It is assumed that a process of giving a quantization value assigned to a quantization representative value whose data value approximates is performed. If the number of quantized representative values is set to 2 ⁿ or less, the output data can be handled by n bits.

[0005] Here, as shown in the figure, linear quantization performs the setting of the representative quantization value equally. On the other hand, when an expected value is obtained in advance, as in the case of predictive coding, the quantization representative is densely close to the expected value and sparsely away from the expected value. It is desirable to perform non-linear quantization to set the value. FIG. 29 shows a case of rounding processing from 3 bits to 2 bits. Against 8 is 2 ^3, by setting the four quantized representative value is 2 ^2, output data can be 2 bits notation.

[0006] In the linear processing, one is simply skipped, a representative quantization value is set, and a quantization value is assigned. For input data having values from 0 to 7, if the value is 0 or 1, the quantization representative value is set to 0 and 2 or 3
If so, the quantized representative value is replaced with a neighboring quantized representative value as in..., And the quantized value assigned to the representative value is given.

On the other hand, in the non-linear processing, if, for example, 3 is an expected value for input data, a quantization representative value is set densely in the vicinity of this value and sparsely as the value is further away from this value. I do. The larger the quantization width, that is, the further apart the quantization representative value, the greater the number of data that can be replaced by the quantization representative value, which means that data with different values is more likely to be treated the same. Therefore,
It can be seen that the closer the expected value is, the better the magnitude relationship of the quantized values reflects the magnitude relationship of the input data.

[0008]

[0005] Non-linear quantization used for predictive coding is performed by various methods, but it is generally difficult to perform it by a simple operation like linear quantization. This is performed by referring to a table such as a ROM table, which leads to an increase in the circuit scale and an increase in processing load, resulting in an increase in cost and a reduction in processing speed.

On the other hand, in predictive coding, since data to be transmitted is a difference value from a predicted value, if an error occurs in the predicted value due to a transmission path error or the like, the error is propagated during reproduction. There was a problem of doing it. For this reason, a method of periodically inserting a PCM value is used for the purpose of keeping such error propagation within a certain range. However, such insertion causes a reduction in compression ratio and unevenness in image quality. It was not a measure.

On the other hand, Japanese Patent Application No. 60-160599 (Japanese Patent Application Laid-Open No.
1390). According to this method, a plurality of nonlinear quantizers are prepared, and a quantizer having a small quantization width in the vicinity of a predicted value is selected and quantized. In this method, the difference is not quantized, but is basically the direct quantization of the input pixel value, so that the error of the predicted value hardly propagates. However, since a plurality of quantizers are provided and switched among them, the circuit scale becomes large, which leads to an increase in cost.

The present invention has been made in view of such circumstances, and has a small circuit scale without using a ROM table.
It is an object of the present invention to provide an image encoding device capable of performing nonlinear quantization at high speed.

Another object of the present invention is to provide an image coding apparatus capable of realizing error propagation in predictive coding by performing high-speed processing with a small circuit scale without reducing the compression ratio. And

Another object of the present invention is to provide an image decoding apparatus capable of performing high-speed decoding processing with a small circuit scale without using a ROM table.

Further, the present invention utilizes the apparatus resources by sharing the above-mentioned small-scale circuits used for quantization and inverse quantization, and uses the apparatus for both encoding and decoding, thereby achieving encoding / decoding. It is an object of the present invention to provide an image encoding / decoding device useful for determining various settings at the time of image conversion.

[0015]

According to a first aspect of the present invention, there is provided an image coding apparatus for converting an input pixel value having a dynamic range of d bits into an n-bit quantized value by encoding. In the image coding apparatus for transmitting the image data, a prediction value generating means for generating a prediction value for an input pixel value from pixels around the input pixel, a quantization width is 2 ^dn in d-bit accuracy, 2 ⁿ obtained by subtracting the preset upper limit number from ⁿ
A linear quantizer generating means for generating a linear quantizer having a linear quantization representative point, and for the linear quantizer, near the periphery of the predicted value, quantization of a number equal to or less than the additional upper limit number A non-linear quantizer generating means for adding a representative point to generate a non-linear quantizer having a quantization width near the predicted value smaller than that of the linear quantizer, and quantizing an input pixel value by the non-linear quantizer. And a quantization means for obtaining a quantization value.

The image encoding apparatus according to a second aspect of the present invention is the image encoding apparatus according to the first aspect, further comprising a shift value generating means for generating a shift value from a predicted value generated by the predicted value generating means, The quantizer generating means, for the linear quantizer generated by the linear quantizer generating means, around the vicinity of the shift predicted value obtained by subtracting the shift value from the predicted value, the additional upper limit number or less The number of quantized representative points is added to generate a non-linear quantizer. The quantizing means calculates a shift input value obtained by subtracting the shift value from an input pixel value in the non-linear quantizer. Is quantized to obtain a quantized value.

According to a third aspect of the present invention, in the image encoding apparatus according to the first aspect, the nonlinear quantizer generating means is arranged so that the non-linear quantizer generates the non-linear quantizer near the periphery of the predicted value with respect to the linear quantizer. From the additional upper limit number, the number of quantization representative points equal to or less than the number obtained by subtracting the preset second additional upper limit number is added, so that the quantization width near the predicted value is smaller than that of the linear quantizer, and The number of quanta equal to or less than the second additional upper limit number is within the range of the additional upper limit number in which the linear quantization representative point is not set by the linear quantizer generator and the quantization width is 2 ^dn. A non-linear quantizer is generated by adding quantization representative points with a quantization width larger than that of the linear quantizer.

According to a fourth aspect of the present invention, there is provided an image coding apparatus for obtaining an n-bit quantized value by encoding from an input pixel value having a dynamic range of d bits and transmitting the quantized value. Prediction value generation means for generating a prediction value for an input pixel value from pixels surrounding the input pixel, and a linear quantization value for generating a linear quantization value by dividing the input pixel value by 2 ^dn with d-bit accuracy Generating means for correcting the linear quantization value based on the difference between the input pixel value and the predicted value to generate a nonlinear quantization value.

According to a fifth aspect of the present invention, in the image encoding apparatus according to the fourth aspect, the image encoding apparatus further comprises a shift value generating means for generating a shift value from the predicted value generated by the predicted value generating means. The quantization value generation means generates a linear quantization value by dividing a shift input value obtained by subtracting the above shift value from the input pixel value by 2 ^dn with d-bit precision.

According to a sixth aspect of the present invention, in the image encoding apparatus according to the fourth aspect, the magnitude of the input pixel value is checked, and if the input pixel value is within a predetermined range, the input pixel value is converted to the linear value. It outputs to the quantized value generating means, and if it is out of the predetermined range, does not output the input pixel value and outputs a predetermined limited quantized value to the non-linear quantized value generating means. Further comprising input value limiting means, wherein the nonlinear quantization value generating means, when the limited quantization value is input from the input value limiting means, uses the limited quantization value as a nonlinear quantization value. It is.

According to a seventh aspect of the present invention, in the image encoding apparatus according to any one of the first to sixth aspects, the linear quantizer generating means or the linear quantized value generating means outputs the predicted value by two. an offset value adding means for determining an offset value according to a remainder obtained by dividing by ^dn and adding the offset value to the position of a linear quantization representative point of the linear quantizer or to the input pixel value It is.

An image coding apparatus according to claim 8 is the apparatus according to any one of claims 1 to 5, wherein the size of the input pixel value is limited to output a limited input value.
It further comprises input value limiting means.

According to a ninth aspect of the present invention, in the image encoding apparatus of the second or fifth aspect, the shift input value limiting means for limiting the magnitude of the shift input value and outputting the limited shift input value is provided. Further provisions.

According to a tenth aspect of the present invention, in the image encoding apparatus according to any one of the first to sixth aspects, the quantizing means or the non-linear quantizing value generating means comprises: Has a quantization value limiting means for limiting the size of.

An image coding apparatus according to an eleventh aspect of the present invention is the image encoding apparatus according to any one of the first to sixth aspects, wherein the quantization means or the non-linear quantization value generation means has a minimum input pixel value. The quantization value is generated such that the quantization value corresponding to the case becomes the minimum value and the quantization value corresponding to the input pixel value increases as the input pixel value increases.

According to a twelfth aspect of the present invention, in the image encoding apparatus according to any one of the first to eleventh aspects, the quantizing means or the non-linear quantized value generating means comprises:
A specific n-bit pattern is defined as an error code and is not used for a quantized value to be output.

According to a thirteenth aspect of the present invention, in the image encoding apparatus according to the eighth or ninth aspect, the quantization means or the nonlinear quantization value generation means converts a specific n-bit pattern into an error code. And is not used for the output quantized value.

According to a fourteenth aspect of the present invention, in the image decoding apparatus for decoding an n-bit quantized value to obtain a reproduced value having a dynamic range of d bits, Prediction value generation means for generating a prediction value with respect to an input quantization value from the pixel of the above, and a number obtained by subtracting a preset additional upper limit number from 2 ⁿ with a quantization width of 2 ^dn in d-bit precision A linear quantizer generating means for generating a linear quantizer having a linear quantized representative point, and the number of quanta less than or equal to the additional upper limit number in the vicinity of the predicted value with respect to the linear quantizer. A non-linear quantizer generating means for generating a non-linear quantizer with a quantization width near the predicted value smaller than that of the linear quantizer by adding a quantization representative point; Dequantize value It is obtained by a inverse quantization means for obtaining a reproduced value each.

The image decoding apparatus according to a fifteenth aspect of the present invention is the image decoding apparatus according to the fourteenth aspect, further comprising a shift value generating means for generating a shift value from the predicted value, wherein the nonlinear quantizer generating means comprises: For the linear quantizer, in the vicinity of a shift predicted value obtained by subtracting the shift value from the predicted value, a quantization representative point equal to or less than the number is added to generate a nonlinear quantizer. It outputs a pixel value obtained by adding the shift value to the reproduction value obtained by the inverse quantization means.

In the image decoding apparatus according to a sixteenth aspect of the present invention, in the apparatus according to the fourteenth aspect, the nonlinear quantizer generating means is arranged such that the non-linear quantizer generating means sets the linear quantizer near the vicinity of the predicted value. From the additional upper limit number, the number of quantization representative points equal to or less than the number obtained by subtracting the preset second additional upper limit number is added, so that the quantization width near the predicted value is smaller than that of the linear quantizer, and The number of quanta equal to or less than the second additional upper limit number is within the range of the additional upper limit number in which the linear quantization representative point is not set by the linear quantizer generator and the quantization width is 2 ^dn. A non-linear quantizer is generated by adding quantization representative points with a quantization width larger than that of the linear quantizer.

According to a seventeenth aspect of the present invention, there is provided an image decoding apparatus for decoding an n-bit quantized value to obtain a reproduced value having a dynamic range of d bits. Predictive value generating means for generating a predictive value for an input quantized value from the pixel of (i), and predictive value linear quantization for generating a predictive value linear quantized value by dividing the predictive value by 2 ^dn with d-bit accuracy Value generation means;
a linear quantization reproduction value generating means for generating a linear quantization reproduction value by multiplying the input quantization value by 2 ^dn in d-bit precision, and a difference between the input quantization value and the prediction value linear quantization value An inverse quantization means for correcting the predicted value or the linearly quantized reproduction value according to a value to generate a reproduction value.

An image decoding apparatus according to an eighteenth aspect of the present invention is the image decoding apparatus according to the seventeenth aspect, further comprising a shift value generating means for generating a shift value from the predicted value, wherein the predicted value linear quantization value generating means is , By dividing the shift predicted value obtained by subtracting the shift value from the predicted value by 2 ^dn ,
This is for generating a predictive value linearly quantized value, and for outputting a pixel value obtained by adding the shift value to the reproduction value generated by the inverse quantization means.

In the image decoding apparatus according to the nineteenth aspect, in the apparatus according to the seventeenth aspect, when the input quantization value is a predetermined value,
A predetermined linear quantized reproduction value is generated.

According to a twentieth aspect of the present invention, there is provided an image coding apparatus according to any one of the fourteenth to nineteenth aspects.
The linear quantizer generating means or the linear quantized reproduction value generating means determines an offset value according to a remainder obtained by dividing the predicted value by ^2dn , and determines a position of a linear quantization representative point of the linear quantizer. Or an offset value adding means for adding the offset value to the linearly quantized reproduction value.

According to a twenty-first aspect of the present invention, in the image decoding apparatus according to the fourteenth aspect,
A specific n-bit pattern is defined as an error code that is not used in normal encoding. If the pattern of the input quantized value matches the defined error code pattern, the predicted value is This is the reproduction value.

The image decoding apparatus according to claim 22 is the image decoding apparatus according to claim 21, wherein when there is a possibility that an error exists in the transmitted input quantization value, the input quantization Replaces the value with the error code defined above.

Further, the image encoding / decoding apparatus according to claim 23 obtains an n-bit quantized value by encoding from an input pixel value having a dynamic range of d bits and transmits the quantized value. Decoding the n-bit quantized value,
In an image encoding / decoding apparatus that obtains a reproduction value having a dynamic range of d bits, prediction that generates a prediction value for an input pixel value or an input quantization value from an input pixel or a pixel around the input quantization value Value generation means and, for d-bit precision, 2
a linear processing value generating means for performing multiplication or division by ^dn to generate a linear processing value, and correcting the linear processing value by a difference value between a predetermined non-linear processing target value and the linear processing value to obtain a non-linear processing value And a non-linear processing value generating means for generating

According to a twenty-fourth aspect of the present invention, in the image encoding / decoding apparatus according to the twenty-third aspect, the image encoding / decoding apparatus further comprises a shift value generating means for generating a shift value from the predicted value. It is a thing.

[0039]

BEST MODE FOR CARRYING OUT THE INVENTION

Embodiment 1 FIG. The image encoding device and the image decoding device according to Embodiment 1 of the present invention realize a nonlinear quantizer by adding a quantization representative point to a linear quantizer having a linear quantization representative value. , Thereby performing quantization or inverse quantization.

FIG. 1 is a block diagram showing a configuration of an image coding apparatus according to Embodiment 1 of the present invention. In the figure, reference numeral 101 denotes a pixel value input unit, which inputs a pixel value having a d-bit dynamic range. Reference numeral 102 denotes a linear quantizer generation unit that generates a linear quantizer having a linear quantization representative point. Numeral 103 denotes a non-linear quantizer generating unit which generates a non-linear quantizer in which a quantization representative point is added in the vicinity of the predicted value and the quantization width near the predicted value is smaller than that of the linear quantizer. . 104 is a quantization unit, which quantizes the input pixel value by the above-mentioned nonlinear quantizer,
Obtain an n-bit quantized value. An output unit 105 outputs the quantization value obtained by the quantization unit. Reference numeral 106 denotes a predicted value generation unit that generates a predicted value for an input pixel value from pixels around the input pixel.

The operation of the image coding apparatus according to the first embodiment configured as described above will be described. Here, for the sake of simplicity, it is assumed that the input pixel value is represented by 8 bits (d) and the quantized value after encoding is 6 bits (n). Here, n = 6 and k = dn = 2. In addition, in order to add the number of quantization representative points, a predetermined number m
= 5.

With respect to the pixel value input from the pixel value input unit 101, the prediction value generation unit 106 uses the pixels around the input pixel as 8-bit linear sums obtained from the quantized values of the surrounding pixels. Generate a predicted value for. Based on the generated predicted values, a linear quantizer generator 102 and a non-linear quantizer generator 103 set quantization representative values for input pixel values.

FIG. 2 is an example of setting the quantization representative value. FIG.
In the example, it is assumed that the predicted value is 13. First, the linear quantizer generator 102 sets a linear quantization representative value including the predicted value 13 as a quantization representative value. The representative value of the linear quantization by the linear quantizer generation unit has a quantization width of 2 ^k
It is assumed that only (2 ⁿ -m) are set. Several meters
Is set in advance according to the number of quantization representative points added in the next processing. In the example of FIG. 2, a quantization representative value having a quantization width of 4 (2 ² ) is set around the prediction value 13. As shown in FIG. 2 (a), within the illustrated range,
About 13, 13, 1, 5, 9 and 17, 21, 25,
29 is set. The number of quantization representative values is set for the whole is 59 from the above set (2 ⁶ -5).

Next, the nonlinear quantizer generator 103 adds m or less quantized representative points around the predicted value and makes the quantization width near the predicted value smaller than that of the linear quantizer. I do. In this example, a total of four (4 <m = 5) quantized representative values are added to a range of two levels before and after the predicted value 13. As shown in FIG. 2B, 11, 12, and 14, 1
5 is added. As a result, the quantization width is reduced to 1 only around the predicted value, and the other portions have the quantization width of 4 as described above. Further, the total number of the quantized representative values after the addition is 63 (<2 ⁶ ), and the quantized value of the encoding result can be represented by 6 bits.

Therefore, the quantization value for the smallest quantized representative value is set to −31, and the smaller quantized value is assigned in order from the smallest quantized representative value. As shown in FIG. 2 (c), in this range, the quantization is performed like the quantization value-2 for the quantization representative value 1... The quantization value 9 for the quantization representative value 29. As a whole, -31
To 31 are assigned.

The quantization unit 104 selects the quantization representative value closest to the input pixel value, and outputs the obtained quantization value to the output unit 105. As shown in FIG. 2D, for example, if the pixel value is 2, the quantization value is -2, and if the pixel value is 14, the quantization value is 14.

As described above, in the image coding apparatus according to the first embodiment, the quantization representative value is basically set by the linear quantization process, and the quantization representative value is added only near the predicted value. Since the quantization process can be performed by an arithmetic process used for linear processing, a ROM table or the like is not required even though the quantization is nonlinear. Also, since the difference between the input pixel value and the predicted value is not processed, but the input pixel value is directly quantized, the magnitude relation between the quantized value representative value (pixel value) and the quantized value matches. For this reason, since each quantized value itself contains absolute level information, even if the predicted value is incorrect, almost no error propagation occurs. In the example of FIG. 2, only an effect occurs only in the portion (nonlinear portion) of the four quantized representative values added around the predicted value.

Next, an image decoding apparatus according to the first embodiment of the present invention, which decodes data encoded as described above, will be described. FIG. 3 is a block diagram illustrating a configuration of the image decoding device according to the first embodiment. In the figure, reference numeral 301 denotes a quantization value input unit for inputting a quantization value as an encoding result. Reference numeral 302 denotes an inverse quantization unit, which is a nonlinear quantizer generated by the nonlinear quantizer generation unit 103, and inversely quantizes an input quantization value. An output unit 303 outputs the result of the decoding. Other codes are the same as those of the image decoding device.

The operation of the image decoding apparatus thus configured will be described below. With respect to the quantized value of the encoding result input from the quantized value input unit 301, the predicted value generating unit 106 calculates a predicted value for the input quantized value as a linear sum from pixels around the input quantized value. Generate. As in the case of the image encoding apparatus, the linear quantization unit generation unit 102 and the non-linear quantization unit generation unit 103 calculate the quantization width around the prediction value based on the prediction value generated by the prediction value generation unit 106. Only a small nonlinear quantizer is generated. The inverse quantization unit 302 inversely quantizes the quantized value input from the quantized value input unit 301 using the quantizer thus formed and outputs the result to the output unit 303.

The image decoding apparatus according to the first embodiment can realize inverse quantization of nonlinear quantization without using a ROM table or the like as in the above-described encoding apparatus. A circuit can be realized.

As described above, in the image coding apparatus and the image decoding apparatus according to the first embodiment, the prediction value generation unit 1
06, a predicted value based on surrounding data is generated for the input data, and the linear quantizer generator 102
And a non-linear quantizer generating unit 103. First, a quantization representative point is set by linear processing, and then a quantization representative value is added around the predicted value. Is to obtain a non-linear quantizer generator with a small value, and perform quantization or inverse quantization by this. Basically, the calculation process used for linear quantization eliminates the need for a ROM table or the like and allows simple addition and subtraction. Since it can be realized by a comparator and a comparator, cost reduction and power saving can be achieved by reducing the circuit scale, and the processing can be speeded up. Also, in the quantization and dequantization processing, since processing is performed on input values, not processing on differences, even if an error occurs in a predicted value, error propagation is minimized without lowering the compression ratio. In a small circuit as described above.

Embodiment 2 The image encoding device and the image decoding device according to the second embodiment of the present invention realize a nonlinear quantizer by adding a quantization representative point to a linear quantizer, as in the first embodiment. To obtain a quantized value, and the shift function prevents the dynamic range from being reduced.

FIG. 4 is a block diagram showing a configuration of an image decoding apparatus according to the second embodiment. In the figure, 10
Reference numeral 7 denotes a shift value generation unit that generates a shift value by a predetermined method. The other parts are the same as those in FIG. 1 and the same as the description in the first embodiment, and the description is omitted here.

The operation of the image coding apparatus according to the second embodiment configured as described above will be described. As in the first embodiment, it is assumed that the input pixel value is represented by 8 bits (d) and the quantized value after encoding is 6 bits (n). Again, n = 6, k = dn = 2. Further, in order to add the number of quantization representative points, it is assumed that m = 5 which is a preset number. For the pixel value input from the pixel value input unit 101, the prediction value generation unit 106 uses pixels around the input pixel as a linear sum obtained from a quantized value of the surrounding pixel,
Generate an 8-bit prediction value. The shift value generation unit 107 generates a shift value obtained by a predetermined method based on the generated prediction value. Based on a shift predicted value obtained by subtracting the shift value from the predicted value, the linear quantizer generator 102 and the nonlinear quantizer generator 103 shift input signals obtained by subtracting the shift value from input pixel values. A quantization representative value for the value is set.

FIG. 5 is a diagram for explaining the setting of a quantization representative point by the apparatus according to the second embodiment. The first embodiment is different from the first embodiment except that a shift predicted value is used instead of a predicted value.
The description is omitted. The quantization unit 104 selects a quantization representative value closest to the shift input value,
The assigned quantization value is output to the output unit 105.

The image coding apparatus according to the second embodiment is also similar to the apparatus according to the first embodiment in that the quantization representative value is basically added only in the vicinity of the shift prediction value using linear quantization. However, no ROM table or the like is required. Further, since the value obtained by shifting the input pixel value is directly quantized, error propagation in predictive coding can be prevented well.

Next, an image decoding apparatus according to the second embodiment of the present invention, which decodes data encoded as described above, will be described. FIG. 6 is a block diagram showing a configuration of the image decoding device according to the second embodiment. In the figure, the shift value generation unit 107 is the same as that of the image encoding device, and the other is the same as that of FIG.

The operation of the image decoding apparatus according to the second embodiment is as follows. For the quantized value of the encoding result input from the quantized value input unit 301, the prediction value generation unit 10
6 generates a prediction value for the input quantization value as a linear sum from pixels around the input quantization value. The shift value generation unit 107 generates a shift value obtained by a predetermined method based on the generated prediction value.

The linear quantizer generator 102 and the non-linear quantizer generator 103 perform nonlinear quantization based on the shift prediction value based on the shift prediction value, in which only the quantization width around the prediction value is small. A vessel is generated. Then, using the obtained nonlinear quantizer, the inverse quantization unit 302 inversely quantizes the quantized value input from the quantized value input unit 301, and outputs a decoding result obtained by adding the shift value to the output unit 303. .

As in the first embodiment, the image decoding apparatus of the second embodiment can realize inverse quantization of nonlinear quantization without using a ROM table or the like. Can be realized.

As described above, in the image coding apparatus and the image decoding apparatus according to the second embodiment of the present invention, the linear quantizer generator 102 and the non-linear quantizer generator 103 perform the same processing as in the first embodiment. Since nonlinear quantization can be performed by arithmetic processing, cost reduction and power saving can be achieved by reducing the circuit scale, and error propagation in predictive coding can be prevented well.

In addition, in the device according to the second embodiment,
A shift value generation unit that generates a shift value according to a predetermined method based on the predicted value generated by the predicted value generation unit, and generates a shift value from the predicted value instead of the predicted value when generating the nonlinear quantizer; By using the shift prediction value obtained by subtracting the value, the effect that the limitation of the dynamic range can be avoided can be obtained.

That is, in the device of the first embodiment, nonlinear quantization can be realized by a small-scale circuit, and error propagation can be prevented by predictive coding. In the setting of, only the number of quantized representative points smaller than the number of quantized representative points to be added next can be set from the number (2 ^k ) that can be originally set.
Then, the setting is performed centering on the prediction point. Therefore, in the dynamic range of the input pixel value, a region having a value larger than the maximum quantized representative value and a region having a value smaller than the minimum quantized representative value are substantially excluded from the target. The range is limited. On the other hand, in the device according to the second embodiment, the limited dynamic range can be recovered by expanding the restricted dynamic range in the direction of the shift by shifting the predicted value by the shift value. , It is possible to realize substantially the same dynamic range.

Embodiment 3 The image encoding device according to the third embodiment of the present invention obtains a nonlinear quantization value by using a linear quantization value generation unit and a nonlinear quantization value generation unit. FIG. 7 is a block diagram illustrating a configuration of an image decoding device according to the third embodiment. In the figure, reference numeral 401 denotes a linear quantization value generation unit, which generates a linear quantization value by performing a division process on an input pixel value. Linear quantization value generator 4
Numeral 01 includes an offset value adding means for determining an offset value based on the remainder of the division process on the predicted value and adding the offset value to the input pixel value. 4
Numeral 02 denotes a non-linear quantized value generating unit which corrects the linear quantized value based on a difference value between the input pixel value and the predicted value to generate a non-linear quantized value. The other parts are the same as those in FIG. 1 and the same as the description in the first embodiment, and the description is omitted here.

The operation of the image coding apparatus according to the third embodiment configured as described above will be described. Pixel value input unit 101
Prediction value generation unit 106
Generates an 8-bit prediction value as a linear sum obtained from quantized values of peripheral pixels using pixels around the input pixel. On the other hand, the input pixel value input from the pixel value input unit 101 is also input to the linear quantization value generation unit 401, and the offset value determined by the lower k bits of the predicted value is added by the offset value adding unit. Then, it is divided by 2 ^k and converted to a linear quantized value. Here, k is obtained as a difference between the number of bits of the dynamic range of the input pixel value and the number of bits of the output quantized value. The linear quantized value is input to the non-linear quantized value generating unit 402, corrected based on the difference between the input pixel value and the predicted value, converted to a non-linear quantized value, and output from the output unit 105.

Here, the algorithm at the time of image encoding will be described. As in the first embodiment, it is assumed that the input pixel value is represented by 8 bits (d) and the quantized value after encoding is 6 bits (n). That is, n = 6, k = dn =
2. FIG. 8 shows the setting of the quantization representative value.
, The four quantization representative values having a quantization width of 1 adjacent to the predicted value (13 in the example of FIG. 8) and the quantization A total of six quantized representative values, including two quantized representative values having a quantization width of 2, are arranged.

FIG. 9 is a flowchart showing an image coding algorithm according to the third embodiment. The operation of the image coding apparatus according to the third embodiment at the time of the quantization value calculation will be described below with reference to the flow of FIG. Here, P (t) represents the predicted value generated at time t, and
(T) is the input pixel value generated at time t, Q (t) is the quantized value generated at time t, RO is the offset value, IQ
Indicates a linear quantization value. Furthermore, for simplicity of explanation,
Here, as the predicted value, the reproduced value at the immediately preceding time is used. The reproduction value is obtained from the predicted value by executing the present algorithm.

In step 1 of the flow of FIG. 9, the offset value adding means of the linear quantization value generation unit 401 uses the lower 2 bits of the predicted value P (−1) obtained from the reproduction value at the immediately preceding time to perform the offset. Set the value. In step 2, after the offset value is added to the input pixel value by the offset value adding means, the result is added to the linear quantized value generation unit 4.
According to 01, it carried out division by 2 ^2, linear quantizing value is determined.

From step 3 onward, the nonlinear quantization value generation unit 4
02 is performed. First, in step 3, a difference value between the predicted value and the input pixel value is obtained, and the sign thereof is determined in step 4, and a correction value is obtained in step 5 or step 6. Based on these, the correction value of the quantization value is calculated in step 7, the reproduction value is calculated in step 8, and in step 9 and step 10, it is determined which value is to be adopted based on the difference value obtained earlier. Then, any one of the steps 10, 12, and 13 is performed to obtain a quantized value. As described above, the conversion from the input pixel value to the quantized value is performed. Similarly, in any of steps 10, 12, and 13, the reproduction value P (0)
Is obtained, and in step 14, the calculation for obtaining the next quantized value is repeated by being fed back and used as the predicted value.

As is clear from this algorithm, the image encoding of the present invention can be executed only by simple addition and subtraction and comparison, and does not require a complicated circuit such as a ROM table despite the use of nonlinear quantization. In addition, in this algorithm, a value almost the same as a value obtained by linearly quantizing an input pixel value is basically obtained as a quantized value, so that the influence of error propagation hardly occurs.

Next, an image decoding apparatus according to the third embodiment of the present invention, which decodes data encoded as described above, will be described. FIG. 10 is a block diagram illustrating a configuration of an image decoding device according to the third embodiment. In the figure, reference numeral 601 denotes a linear reproduction value generation unit, which generates a linear quantization reproduction value for an input quantization value. The linear reproduction value generation unit 601 includes an offset value adding unit that determines an offset value based on the remainder of the division process on the predicted value, and adds the offset value to the linear reproduction value. 602 is a non-linear quantized reproduction value generation unit,
Correction is performed based on the difference between the input quantization value and the predicted value linear quantization value. Reference numeral 603 denotes a prediction value linear quantization value generation unit that obtains a prediction value linear quantization value by performing a division process on the prediction value. The quantization value input unit 301, the output unit 303, and the predicted value generation unit 106 are the same as those in FIG.

The operation of the image decoding apparatus according to the third embodiment is as follows. The offset value adding means of the linear reproduction value generation unit 601 obtains a predetermined offset value for the lower k bits of the predicted value generated by the predicted value generation unit 106. Next, the linear reproduction value generation unit 601 performs a multiplication process on the input quantization value input from the quantization value input unit 301, and subtracts the previous offset value from the resulting (2 ^k ) -multiplied value. The values are added, converted to a linearly quantized reproduction value, and output to the non-linear quantization reproduction value generation unit 602.

On the other hand, the prediction value linear quantization value generation unit 602
Divides the predicted value by (2 ^k ), obtains a linear quantized value of the predicted value, and outputs this to the nonlinear quantized reproduction value generation unit 602. The non-linear quantization reproduction value generation unit 602 adds a correction value generated by a difference value between the input quantization value and the prediction value linear quantization value to the linear quantization reproduction value or the prediction value, and obtains the obtained quantization. The reproduction value is output to the output unit 303.

Here, the algorithm of image decoding will be described below. The input pixel value and the reproduction value at the time of encoding are represented by 8 bits (d), and the reproduction value obtained as a result of decoding is also 8 bits. The quantized value after encoding is 6 bits (n) (n = 6, k = dn = 2). Also,
As for the setting of the quantization representative value, as shown in FIG. 8, for the linear quantization representative value whose quantization width is 4, four quantization widths 1 adjacent to the predicted value (13 in the example of FIG. 8) are used. , And two quantized representative values having a quantization width of 2, that is, a total of six quantized representative values are arranged.

FIG. 11 is a flowchart showing an image decoding algorithm according to the third embodiment. The algorithm that the image encoding apparatus according to the present embodiment performs in the operation of the non-linear quantization value according to the flow of FIG. 11 will be described below. Here, P (t) represents a quantized reproduction value generated at time t and a predicted value, Q (t) represents an input quantized value input at time t, RO represents an offset value, and PQ represents a predicted value. The linear quantization value, IQ indicates a linear reproduction value. For the sake of simplicity, the quantized reproduction value at the immediately preceding time is used as the prediction value.

In step 1 of the flow of FIG. 11, the offset value adding means of the linear reproduction value generation unit 601 uses the lower 2 bits of the predicted value P (−1) obtained from the quantized reproduction value at the immediately preceding time. Set the offset value. Step 2
Then, for the predicted value, the predicted value linear quantization value generation unit 60
The division by 3 and the division by 2 ² are performed, and the predicted value linear quantization value PQ is obtained.

After step 3, correction is performed by the nonlinear quantized reproduction value generation unit 602. First, in step 3, a difference value between the predicted value linear quantization value and the input quantization value is obtained, and the sign thereof is determined in step 4, and a correction value is obtained in step 5 or step 6. After calculating the corrected reproduction value in Step 7 based on these, in Steps 8 and 10 it is determined which value is to be adopted based on the difference value obtained previously, and Steps 9, 11, and
By performing any one of the steps 12, a quantized reproduction value is obtained. As described above, the decoding process from the input quantized value to the quantized reproduced value is performed, and the quantized reproduced value is fed back in step 13 to
The calculation for obtaining the next quantized reproduction value is repeated.

Although the above-mentioned image decoding algorithm also uses nonlinear quantization, RO for inverse quantization is used.
It can be realized only by simple addition and subtraction and comparison without using an M table or the like. Therefore, it can be realized with a very small circuit although the quantization representative value adaptively changes depending on the predicted value.

As described above, in the image coding apparatus according to the third embodiment, the predicted value generation unit 106 generates a predicted value based on surrounding data for input data, and generates the linear quantized value generation unit. 401, a non-linear quantized value generation unit 402, and first sets a quantized representative point by linear processing, and then adds a quantized representative value around the predicted value, so that quantization is performed only around the predicted value. A quantization value is obtained by performing a non-linear quantization process having a small width. Therefore, since the calculation is basically performed by using the linear quantization, a ROM table or the like is not required,
Since it can be realized by a simple adder / subtracter and a comparator, it is possible to reduce the cost and power consumption by reducing the circuit scale, and also to increase the processing speed.

The image decoding apparatus according to the third embodiment includes a linear reproduction value generation unit 601, obtains a linear quantized reproduction value by linear processing, and uses the prediction value generated by the prediction value generation unit 106. , Predicted value linear quantization value generation unit 603
Obtains a prediction value linear quantization value, and based on these, the nonlinear quantization reproduction value generation unit 602 performs correction by a difference value between the input quantization value and the prediction value linear quantization value to obtain a quantization reproduction value. By doing so, inverse quantization in a small-scale circuit can be realized, and cost reduction and power saving can be achieved.

In the quantization and dequantization processing, since the processing is performed not on the difference but on the input value, even if an error occurs in the predicted value, the error propagation is performed without lowering the compression ratio. Can be realized in a small circuit as described above.

Embodiment 4 The image coding apparatus according to the fourth embodiment of the present invention obtains a non-linear quantized value by a linear quantized value generating unit and a non-linear quantized value generating unit, as in the third embodiment. To prevent the dynamic range from being reduced. FIG. 12 shows Embodiment 4
FIG. 3 is a block diagram illustrating a configuration of an image decoding device according to the first embodiment.
In the figure, reference numeral 107 denotes a shift value generation unit, which generates a shift value by a predetermined method. Others are the same as FIG.
Since the description is the same as that in the third embodiment, the description is omitted here.

The operation of the image coding apparatus according to Embodiment 4 configured as described above will be described. Pixel value input unit 101
Prediction value generation unit 106
Generates an 8-bit prediction value as a linear sum obtained from quantized values of peripheral pixels using pixels around the input pixel. Shift value generating section 107 based on the generated predicted value
Generates a shift value obtained by a predetermined method.
On the other hand, the shift input value obtained by subtracting the shift value from the input pixel value input from the pixel value input unit 101 is input to the linear quantization value generation unit 401, and is determined by the lower k bits of the predicted value by the offset value adding unit. After the offset value is added, the result is divided by 2 ^k and converted to a linear quantized value. Here, k is obtained as a difference between the number of bits of the dynamic range of the input pixel value and the number of bits of the output quantized value.

The linear quantized value is calculated by the non-linear quantized value generating section 4.
02, is corrected based on the difference between the shift input value and the shift prediction value, is converted to a non-linear quantization value, and is output from the output unit 105.

Here, the algorithm of image coding will be described. As in the third embodiment, it is assumed that the input pixel value is represented by 8 bits (d) and the quantized value after encoding is 6 bits (n). That is, n = 6, k = dn =
2. Further, as for the setting of the quantization representative value, as in the third embodiment, a total of 6 quantization representative values of four quantization representative values of quantization width 1 and two quantization representative values of quantization width 2 is used. It is assumed that the quantization representative values are additionally arranged. However, a shift prediction value is used instead of the prediction value.

FIG. 13 is a flowchart showing an image encoding algorithm according to the fourth embodiment. The algorithm that the image encoding apparatus according to the present embodiment performs in the operation of the nonlinear quantization value calculation according to the flow of FIG. 13 will be described below. Here, P (t) represents the predicted value generated at time t, Sf is the shift value obtained from the predicted value, I (t) is the input pixel value generated at time t, Q
(T) indicates a quantization value generated at time t, RO indicates an offset value, and IQ indicates a linear quantization value. Further, for simplicity of description, here, the reproduction value at the immediately preceding time is used as the prediction value. The reproduction value is obtained from the predicted value by executing the algorithm.

In step 1 of the flow shown in FIG. 13, the shift value generation unit 107 calculates the predicted value P obtained from the reproduction value at the immediately preceding time.
The shift value Sf is determined by the lower 5 bits of (-1). Then, in step 2, a shift input value obtained by subtracting the shift value from the input pixel value and a shift predicted value obtained by subtracting the shift value from the predicted value are obtained. In step 3, the offset value adding means of the linear quantization value generation unit 401 sets an offset value using the lower two bits of the shift prediction value. In step 4, the offset value adding means, after the addition of the offset value to the shift input value, due to the linear quantization value generating unit 401 to this effect, is performed division by 2 ^2, linear quantizing value is determined .

After step 5, the non-linear quantization value generating section 4
02 is performed. First, in step 5, a difference value between the shift predicted value and the shift input value is obtained, and the sign thereof is determined in step 6, and step 7 or step 8 is performed.
Is used to determine the correction value. Based on these, the correction value of the quantization value is calculated in steps 8 and 9, and the reproduction value is calculated in step 10, and in step 11 and step 13, any value is adopted based on the difference value obtained earlier. Is determined, and one of Steps 12, 14, and 15 is executed to obtain a quantized value.

As described above, the conversion from the input pixel value to the quantized value is performed.
In any one of 15, the reproduction value P (0) to which the shift value has been added is obtained, and in step 16, the feedback is used as a prediction value, and the calculation for obtaining the next quantization value is repeated.

As is clear from this algorithm, the image encoding of the present invention can be executed only by simple addition and subtraction and comparison, and does not require a complicated circuit such as a ROM table despite the use of nonlinear quantization. In addition, in this algorithm, a value almost the same as a value obtained by linearly quantizing an input pixel value is basically obtained as a quantized value, so that the influence of error propagation hardly occurs.

Next, an image decoding apparatus according to the fourth embodiment of the present invention, which decodes data encoded as described above, will be described. FIG. 14 is a block diagram illustrating a configuration of an image decoding device according to the third embodiment. In the figure, reference numeral 107 denotes a shift value generation unit, which generates a shift value by a predetermined method. The other parts are the same as those in FIG. 10 and are the same as the description in the third embodiment, and thus the description is omitted here.

The operation of the image decoding apparatus according to the fourth embodiment is as follows. The shift value generation unit 107 generates a shift value obtained by a predetermined method based on the prediction value generated by the prediction value generation unit 106. Linear reproduction value generation unit 601
The offset value adding means has a predetermined offset value based on the lower k bits of the shift predicted value obtained by subtracting the shift value from the predicted value. Next, the linear reproduction value generation unit 601 performs a multiplication process on the input quantization value input from the quantization value input unit 301, and obtains the result (2 ^k ).
The offset value is added to the multiplied value, converted to a linear quantized reproduction value, and converted to a non-linear quantized reproduction value generation unit 602.
Output to

On the other hand, the prediction value linear quantization value generation section 602
Divides the shift prediction value by (2 ^k ) to obtain a prediction value linear quantization value, and outputs this to the nonlinear quantization reproduction value generation unit 602. The non-linear quantization reproduction value generation unit 602 adds a correction value generated by a difference value between the input quantization value and the prediction value linear quantization value to the linear quantization reproduction value or the prediction value, and obtains the obtained quantization. The shift value is added to the reproduction value and output to the output unit 303.

Here, the algorithm of image decoding will be described below. The input pixel value and the reproduction value at the time of encoding are represented by 8 bits (d), and the quantized value after encoding is 6 bits (n) (n = 6, k = d− n = 2).
As for the setting of the quantization representative value, similarly to the case of the image coding apparatus, four quantization representative values having a quantization width of 1 and
A total of six quantized representative values including two quantized representative values having a quantization width of 2 are additionally arranged.

FIG. 15 is a flowchart showing an image decoding algorithm according to the fourth embodiment. Hereinafter, an algorithm that the image encoding device according to the present embodiment performs in the operation of the non-linear quantization value according to the flow of FIG. 15 will be described. Here, P (t) represents a quantized reproduction value generated at time t and a predicted value, Q (t) represents an input quantized value input at time t, RO represents an offset value, and PQ represents a predicted value. The linear quantization value, Sf indicates a shift value obtained from the predicted value, and IQ indicates a linear reproduction value. For the sake of simplicity, the reproduction value at the immediately preceding time is used as the prediction value.

In step 1 of the flow of FIG. 15, the shift value generation unit 107 calculates the prediction value P obtained from the reproduction value at the immediately preceding time.
The shift value Sf is determined by the lower 5 bits of (-1). Then, in step 2, a shift predicted value obtained by subtracting the shift value from the predicted value is obtained. In step 3, the offset value adding means of the linear reproduction value generation unit 601 sets an offset value using the lower two bits of the predicted value P (-1) obtained from the quantized reproduction value at the immediately preceding time. In step 4, division by 2 ² is performed on the shift prediction value by the prediction value linear quantization value generation unit 603 to obtain a prediction value linear quantization value PQ.

After step 5, correction is performed by the nonlinear quantized reproduction value generation unit 602. First, in step 5, a difference value between the predicted value linear quantization value and the input quantization value is obtained, and the sign thereof is determined in step 6, and a correction value is obtained in step 7 or step 8. After calculating the corrected reproduction value in Step 9 based on these, in Steps 10 and 12, it is determined which value is to be adopted based on the difference value obtained previously, and Steps 11 and 1 are performed.
By performing any one of the steps 3 and 14, a quantized reproduction value is obtained, and the shift value is added and output.

As described above, the decoding process from the input quantized value to the reproduced value is performed, and the reproduced value is fed back in step 15, and the operation for obtaining the next quantized reproduced value is repeated.

Although the above-mentioned image decoding algorithm also uses nonlinear quantization, RO for inverse quantization is used.
It can be realized only by simple addition and subtraction and comparison without using an M table or the like. Therefore, it can be realized with a very small circuit although the quantization representative value adaptively changes depending on the predicted value.

As described above, in the image coding apparatus according to the fourth embodiment, the predicted value generation section 106 generates a predicted value based on surrounding data for input data, and generates the linear quantized value generation section. 401, a non-linear quantized value generation unit 402, and first sets a quantized representative point by linear processing, and then adds a quantized representative value around the predicted value, so that quantization is performed only around the predicted value. A quantization value is obtained by performing a non-linear quantization process having a small width. Therefore, as in the third embodiment, by realizing nonlinear processing with a small circuit scale, cost reduction, power saving, and quick processing are possible, and error propagation in predictive coding can be prevented well.

The image decoding apparatus according to the fourth embodiment includes a linear reproduction value generation unit 601, obtains a linear quantization reproduction value by linear processing, and obtains a linear quantization reproduction value based on the prediction value generated by the prediction value generation unit 106. , Predicted value linear quantization value generation unit 603
Obtains the predicted linear quantization values, and based on these, the non-linear quantization reproduction value generation unit 602 performs correction by the difference between the input quantization value and the prediction linear quantization values to obtain the quantization reproduction values. As a result, as in the third embodiment, inverse quantization in a small-scale circuit can be realized, and cost reduction and power saving can be achieved.

In addition, in the device according to the fourth embodiment,
A shift value generation unit that generates a shift value according to a predetermined method based on the predicted value generated by the predicted value generation unit, and generates a shift value from the predicted value instead of the predicted value when generating the nonlinear quantizer; By using the shift prediction value obtained by subtracting the value, the dynamic range limited by the decrease in the linear quantization representative value can be recovered by expanding in the shift direction, similar to the second embodiment. In addition, it is possible to realize substantially the same dynamic range as when linear quantization is exclusively performed.
It should be noted that the correction at the time of calculating the nonlinear quantization value or the nonlinear quantization reproduction value is performed based on the difference value between the shift input value and the shift prediction value. And the same effect can be obtained.

Embodiment 5 FIG. The image coding apparatus and the image decoding apparatus according to the fifth embodiment of the present invention realize a nonlinear quantizer by adding a quantization representative point to a linear quantizer, as in the first embodiment. , And has a function of setting a quantization representative point in a specific area.

FIG. 16 is a block diagram showing a configuration of an image decoding apparatus according to the fifth embodiment. A non-linear quantizer generator 801 generates a non-linear quantizer in which a quantization representative point is added to a specific region in addition to the vicinity of a prediction point. The other parts are the same as those in FIG. 1 and the same as the description in the first embodiment, and the description is omitted here.

The operation of the image coding apparatus according to the second embodiment configured as described above will be described. As in the first embodiment, it is assumed that the input pixel value is represented by 8 bits (d) and the quantized value after encoding is 6 bits (n). Again, n = 6, k = dn = 2. In addition, in order to set the number of quantization representative points to be added, a predetermined number m =
5 is assumed. Further, it is assumed that p = 3, which is a preset number, for adding a quantization representative point to a specific area.

From the pixel value input, the linear quantizer generator 10
2 is the same as that of the first embodiment up to the generation of the linear quantizer having the linear quantization representative value. The representative value of the linear quantization by the linear quantizer generation unit 102 has a quantization width of 2
It becomes 4 which is ^2, which is 59 (2 ⁶ -5).

FIG. 17 is a diagram for explaining the setting of the quantization representative value in the image coding apparatus according to the fifth embodiment. FIG. 17A shows that the linear quantizer generator 102
It is a figure showing the state of quantization representative value setting. As shown in the figure, in this case, the region from -128 to -119 and 119
To 127 are outside the region selected as the quantization representative value.

Next, the non-linear quantizer generator 801 adds a quantization representative value to the linear quantization representative value. First, mp or less quantized representative points are added around the predicted value to make the quantization width near the predicted value smaller than that of the linear quantizer. In this example, 2 before and after the predicted value 13
A total of two (mp = 2) quantized representative values are added to the level range. As shown in FIG.
Is added. As a result, the quantization width is reduced to 2 only around the predicted value, and the other portions are 4 as described above.
Is obtained. The total number of quantization representative points is 61.

Next, the non-linear quantizer generator 801 calculates
One quantization representative value is additionally set for each of the regions 128 to -119 and the regions 119 to 127 (2 <p = 3). For each area, the quantization width is 8, the total number of quantized representative values after addition is 63 (<2 ⁶ ), and the quantized value of the encoding result can be represented by 6 bits.

As shown in FIG. 17 (c), the quantization value is allocated to the minimum -31 for the region from -128 to -119, and the quantization representative value is assigned in the same manner as in the first embodiment. Quantization values are assigned in ascending order of
31 is allocated to the area from 9 to 127. The quantization by the quantization unit 104 and the output to the output unit 108 are performed in the same manner as in the first embodiment.

Next, an image decoding apparatus according to the fifth embodiment of the present invention for decoding data encoded as described above will be described. FIG. 18 is a block diagram illustrating a configuration of an image decoding device according to the fifth embodiment. In the figure, the nonlinear quantizer generation unit 901 has a function of adding a quantization representative value to a specific area, similarly to the image coding apparatus, and the other components are the same as those in FIG. Omitted.

In the operation of the image decoder according to the fifth embodiment, the representative quantization value of the nonlinear quantizer generated by the nonlinear quantizer generator 901 is as shown in FIG. Except for this point, the image decoding apparatus is the same as the image decoding apparatus according to the first embodiment. The image decoding apparatus according to the fifth embodiment also can realize the decoding process in a small-scale circuit without requiring a ROM table or the like.

As described above, the image coding apparatus and the image decoding apparatus according to the fifth embodiment include the nonlinear quantizer generator having the function of adding the quantization representative value to the specific area. The limitation of the dynamic range due to the addition of the quantized representative value in the vicinity of the predicted value can be reduced even by the configuration having no shift value generation function in the device of the second embodiment.

Embodiment 6 FIG. The image coding apparatus and the image decoding apparatus according to the sixth embodiment of the present invention obtain a non-linear quantization value by a linear quantization value generation unit and a non-linear quantization value generation unit, as in the second embodiment. It has a function of performing a predetermined encoding or decoding on an input of a specific area.

FIG. 19 is a block diagram showing a configuration of an image decoding apparatus according to the sixth embodiment. An input limiting unit 1001 checks the size of an input pixel value input by the pixel value input unit 101 and outputs the input pixel value to a linear quantization value generation unit 401 when the input pixel value is within a predetermined range. If it is outside, the input pixel value is not output and a predetermined quantization value is output to the nonlinear quantization value generation unit 1002. Reference numeral 1002 denotes a non-linear quantized value generation unit. When a quantized value is output from the input value limiting unit 1001, it is output to the output unit 105 as a non-linear quantized value. Other configurations are the same as those in FIG. 7 and are the same as those described in the third embodiment, and thus description thereof is omitted here.

The operation of the image coding apparatus according to the sixth embodiment having the above configuration will be described. As in the first embodiment, it is assumed that the input pixel value is represented by 8 bits (d) and the quantized value after encoding is 6 bits (n). Again, n = 6, k = dn = 2. In addition, in order to set the number of quantization representative points to be added, a predetermined number m =
5 is assumed. Further, it is assumed that p = 3, which is a preset number, for adding a quantization representative point to a specific area.

For a pixel value input from the pixel value input unit 101, an 8-bit prediction value is generated by the prediction value generation unit 106 using pixels around the input pixel. The input pixel value input from the pixel value input unit 101 is input to the input value control unit 1001 to determine whether the input pixel value is within a predetermined area.

In this determination, if the value is within a predetermined area, the value as it is is input to linear quantization section 401, and linear quantization value generation section 401 adds an offset value determined by the lower k bits of the predicted value. After that, as in the third embodiment, the data is converted into a linear quantized value by a division process. As in the third embodiment, the linear quantization value is corrected by the nonlinear quantization value generation unit 1002 based on the difference value between the input value and the predicted value, and is converted into a nonlinear quantization value.
Output.

On the other hand, when the input value is out of the predetermined area in the judgment by the input value restriction unit 1001, the input value restriction unit 1001 generates a non-linear quantized value by a predetermined method.
This nonlinear quantization value is input to the nonlinear quantization value generation unit 1002, and is output from the output unit 105.

FIG. 21 is a diagram for describing encoding and decoding according to the sixth embodiment. As shown in FIG. 21A, the input value limiting unit 1001 determines that the input pixel value
If the number is between 18, the linear quantization value generation unit 40
1, and the output input pixel value is processed by the linear quantization value generation unit 401 and the non-linear quantization value generation unit 1002 in the same manner as in the encoding device of the third embodiment, and the quantization value is can get.

On the other hand, if the input pixel value is -128 to -119,
Alternatively, if the value is between 119 and 127, the input value limiting unit 1002 does not output to the linear quantization value generation unit 401, generates a non-linear quantization value “−31” or “31”, This is output to the nonlinear quantization value generation unit 1002.
As a result, an encoding result similar to that of the fifth embodiment can be obtained.

Next, an image decoding apparatus according to the sixth embodiment of the present invention for decoding data encoded as described above will be described. FIG. 20 is a block diagram illustrating a configuration of an image decoding device according to the sixth embodiment. In the figure, when a specific quantized value is input, a non-linear quantized reproduced value generating unit 1201 outputs a predetermined reproduced value.

When the quantized value input to the quantized value input unit 301 is a specific value, the quantized value input unit 301 outputs the quantized value to the non-linear quantized reproduction value generating unit 1201 and outputs the non-linear quantized value. The generalized reproduction value generation unit 1201 outputs a predetermined reproduction value to the output unit 308. Here, the specific thing is
This is a predetermined nonlinear quantization value output from the input value limiting unit 1001 to the nonlinear quantization value generation unit 1002 in the image encoding device. Except for this case, the processing is performed by the same operation as that of the image decoding apparatus according to the third embodiment.

As shown in FIG. 21 (b), when the quantized value input to the quantized value input unit is other than -31 or 31, processing is performed in the same manner as in the third embodiment, and the reproduced value is changed. can get. On the other hand, if the input quantization value is -31 or 31
If, the nonlinear quantized reproduction value generation unit 1201
The reproduction values of -123 and 123 are output to the output unit 30 respectively.
8 is output. As a result, the same decoding result as in the fifth embodiment can be obtained.

The image encoding device and the image decoding device according to the sixth embodiment also do not require a ROM table or the like.
Encoding / decoding overprocessing can be realized in a small-scale circuit. Further, by having a function of outputting a predetermined numerical value for an input pixel value of a specific area, or a quantization value,
The limitation of the dynamic range due to the addition of the quantized representative value in the vicinity of the predicted value can be reduced even by the configuration having no shift value generation function in the device of the second embodiment.

Embodiment 7 FIG. The image coding apparatus according to Embodiment 7 of the present invention has an input restriction function. In the image coding device according to the first or third embodiment, the image coding device realizes nonlinear quantization by adding a quantization representative value near a predicted value to a quantization representative value obtained by linear quantization. For this reason, the number of linear quantization representative values decreases by the number of added quantization representative values, and it is necessary to limit the quantization range (dynamic range) in linear quantization.

FIG. 22 is a block diagram showing a configuration of an image coding apparatus according to Embodiment 7 of the present invention. 22 in FIG.
Reference numeral 01 denotes an input value limiting unit that limits an input value. 702
Denotes an image encoding unit, which includes the linear quantizer generation unit, the non-linear quantizer generation unit, the quantization unit, and the prediction value generation unit described in the first embodiment. Pixel value input unit 101 and output unit 1
05 is the same as FIG.

[0129] Encoding of the image encoding apparatus according to the seventh embodiment is performed by the following operation. First, the input pixel value input from the pixel value input unit 101 in FIG. 22 is first input to the input value limiting unit 701, and is limited to a range that can be quantized by the linear quantizer of the image encoding unit 702. The restricted input value is sent to the image encoding unit 702 in the first embodiment.
Is converted into a quantized value by the same encoding processing as described above, and the result is output to the output unit 105. Input value limiter 701
When adding m quantized representative points in the non-linear quantizer, it is necessary to limit the dynamic range of the input pixel value by at least the level of m × 2 ^k .

As described above, the image coding apparatus according to the seventh embodiment includes input value limiting section 701, and limits the input pixel value to be input to a range that can be processed by linear quantization processing. As in the first embodiment, in the non-linear quantization processing that prevents error propagation in a small-scale circuit, it is possible to well retain the characteristic of good image quality due to predictive coding. In the seventh embodiment, the configuration of the image encoding unit 702 conforms to the configuration of the first embodiment. However, the configuration of the image encoding unit 702 may conform to the configuration of the third embodiment. can get.

Embodiment 8 FIG. The image encoding device according to the eighth embodiment of the present invention has a quantization value limiting function. FIG. 23 is a block diagram showing a configuration of an image encoding device according to Embodiment 8 of the present invention. Reference numeral 703 in FIG. 23 denotes a quantization value limiting unit that limits the quantization value. Others are in Figure 2
Same as 2.

The coding by the image coding apparatus according to the eighth embodiment is performed by the following operation. First, the input pixel value input from the pixel value input unit 101 in FIG. 22 is converted into a quantized value in the image encoding unit 702 by the same encoding processing as in the first embodiment, and then the quantized value limiting unit 703 is input. The quantization value limiting unit 703 detects the value of the input quantization value, and if it exceeds the range from the minimum value to the maximum value determined for the quantization value, limits the quantization value to the range and sends it to the output unit 105. Output.

As described above, the image coding apparatus according to the eighth embodiment includes the quantization value limiting section 703, and limits the quantization value obtained as a result of the coding to a certain range. Eliminate dynamic range limitations,
In addition, it is possible to maintain the characteristic of good image quality in predictive coding in non-linear quantization processing that prevents error propagation in a small-scale circuit.

In the eighth embodiment, the configuration of the image encoding unit 702 is similar to that of the first embodiment. However, the configuration of the image encoding unit 702 may be similar to that of the third embodiment. The effect of is obtained.

Embodiment 9 FIG. The image coding apparatus according to Embodiment 9 of the present invention has a shift input restriction function. The configuration of the image coding apparatus according to the ninth embodiment is the same as that of the fourth embodiment, and FIG. 6 is used for the description. Here, the linear quantization value generation unit 102 includes a shift input value limiting unit that limits a shift input value obtained from an input pixel value and a shift value to a certain range. Others are the same as the fourth embodiment.

FIG. 24 is a flowchart showing an encoding algorithm of the image encoding apparatus according to the ninth embodiment. The flow of FIG. 24 differs from the flow of FIG. 13 only in that the range of the shift input value is limited in step 2. Except for this point, the device according to the ninth embodiment operates the same as the device according to the fourth embodiment.

As described above, the image coding apparatus according to the ninth embodiment has the same effect as that of the seventh embodiment by providing the shift input restriction function.
In the device according to the seventh embodiment, the input data excluded from the target by the input restricting unit can be restricted according to the setting of the shift value by restricting the shift input as compared with the case where the input data is not used at all. Therefore, more flexible utilization is possible. Although the ninth embodiment is based on the fourth embodiment, the ninth embodiment can be based on the second embodiment, and the same effects can be obtained.

Embodiment 10 FIG. Embodiment 10 of the present invention
The image coding apparatus and the image decoding apparatus according to the above-mentioned are capable of setting an error code. The configuration of the image coding apparatus according to the tenth embodiment is the same as that of the eighth embodiment, and FIG. 23 is used for the description. In the image encoding apparatus according to the tenth embodiment configured as described above, the quantization value limiting unit 11
By having 01, it becomes possible to prohibit use of a predetermined n-bit pattern by normal quantization. Further, this makes it possible to assign an error code indicating that an error has occurred in the quantized value to the pattern whose use is prohibited. For example, if the quantization value is 6
If it is a bit, -31 to 31 can be used for quantization, and -32 can be set to the error code.

FIG. 25 is a block diagram showing a configuration of an image decoding apparatus according to the tenth embodiment capable of performing error code processing. An image decoding unit 904 in FIG. 25 includes the linear quantizer generation unit, the non-linear quantizer generation unit, and the prediction value generation unit illustrated in FIG. 3, and is similar to the image decoding device according to the first embodiment. Is performed. An error code detection unit 902 detects an error code with respect to the input quantization value. A prediction error setting unit 903 sets a difference from a prediction value.

The decoding processing by the decoding apparatus according to the tenth embodiment is performed by the following operation. First, an error code detection unit 902 determines whether or not an input quantization value input from the quantization value input unit 301 in FIG. 25 is an error code. Here, if it is an error code, the image decoding unit 904 receives the error code via the prediction error setting unit 903.
By setting the difference from the predicted value to 0, the predicted value is output to the output unit 303 as it is. When the error code detection unit 902 determines that the input quantized value is not an error code, the image decoding unit 904 performs normal decoding on the input quantized value, and outputs the result to the output unit. 303
Output to When an error code is detected by the above processing, a predicted value is output as a reproduced value.

By replacing an erroneous portion in the data encoded by the image encoding device with an error code, the above-described function of the decoding device becomes effective. The error code insertion function will be described.

FIG. 26 is a block diagram showing an example of a circuit for inserting an error code. Reference numeral 1201 in FIG. 26 denotes a reproduction value input unit, which reproduces encoded data. An error code replacement unit 1202 replaces a quantized value of the reproduced data with an error code. An error detection unit 1203 detects an error in the reproduced data. An output unit 1204 outputs the data after the replacement.

In the error code insertion circuit configured as described above, data reproduced from a magnetic tape or a transmission signal by the reproduction value input unit 1201 is detected by the error detection unit 1203 to determine the presence or absence of an error. Here, the quantized value where an error is detected is replaced with an error code by an error code replacement unit 1202. In this way, the quantized value in which an error may have occurred is replaced with the error code and output to the output unit 1204.

As described above, in the image coding apparatus and the image decoding apparatus according to the tenth embodiment, the image coding apparatus includes the quantization value input unit 301 and sets a specific quantization value as an error code. The obtained coded data is subjected to error code replacement processing by the error code insertion circuit as described above, so that the error code detection unit 902 and the prediction error setting unit 90
In the case where an error code is detected in the image decoding apparatus provided with No. 3, the prediction value is used as it is as the reproduction value, so that the effect of the error can be minimized.

Embodiment 11 FIG. Embodiment 11 of the present invention
Is an image encoding / decoding device capable of recording and reproducing an image. Many of the processes common to the image encoding device and the image decoding device of the present invention described in the above embodiment are common. . Therefore, in the eleventh embodiment, the common portion is shared, and an apparatus capable of performing both encoding and decoding is provided.

FIG. 27 is a block diagram showing a configuration of an image encoding / decoding device according to the eleventh embodiment. In the figure, reference numeral 1051 denotes a pixel value input unit for inputting a pixel value having a d-bit dynamic range. Reference numeral 1052 denotes a quantization value input unit, which inputs a quantization value that is an encoding result. An input switching unit 1053 switches an input to be processed in the apparatus. Reference numeral 1054 denotes a quantization / inverse quantization processing unit that performs a quantization process in encoding and an inverse quantization process in decoding. The quantization / inverse quantization processing unit 1054 includes a predicted value generation unit 1055, a linear processing unit 1056, and a non-linear processing unit 1057. The predicted value generation unit 1055 generates a predicted value for the input. The linear processing means 1056 obtains a linear processing value by performing a division process or a multiplication process on the input value. The non-linear processing means 1057 obtains a non-linear processing value from the difference between the linear processing value and the input value. Reference numeral 1058 denotes an output switching unit that switches the output destination of the nonlinear processing value output from the quantization / inverse quantization processing unit 1054. Reference numeral 1059 denotes a quantization value output unit, which outputs a quantization value as a result of encoding. A reproduction value output unit 1060 outputs a reproduction value as a decoding result.

The eleventh embodiment configured as described above
The operation of the image encoding / decoding device according to the first embodiment at the time of encoding and decoding will be described below. First, during the encoding process,
The input switching unit 1053 includes a quantization / inverse quantization processing unit 105
4 is set as an input from the pixel value input unit 1051, and the output switching unit 1058 sets the input from the quantization / inverse quantization processing unit 1054 to the quantization value output unit 1051.
9 is set.

When the pixel value is input to the pixel value input unit 1051, the input pixel value is quantized and dequantized by the quantization unit 1054.
And the predicted value generating means 1055 generates a predicted value based on the input pixel value. Thereafter, the quantization / inverse quantization processing unit 1054 functions as the prediction value generation unit 106, the linear quantization value generation unit 401, and the nonlinear quantization value generation unit 402 of the image encoding device according to the third embodiment (number Performs encoding processing according to the algorithm of FIG. 7) and FIG.
The obtained quantization value is output to the quantization value output unit 1059.

Next, in the case of the decoding process, the input switching unit 105
3 sets an output to the quantization / inverse quantization processing unit 1054 as an input from the quantization value input unit 1052;
The output switching unit 1058 includes a quantization / inverse quantization processing unit 105
4 is set to be output to the reproduction value output unit 1060.

When the quantized value is input from the quantized value input unit 1052 to the quantization / dequantization processing unit 1054, the quantization / dequantization processing unit 1054 operates in the image decoding apparatus according to the third embodiment. Predicted value generator 106, linear reproduction value generator 6
01, and functions as a nonlinear quantized reproduction value generation unit 602 and a predicted value linear quantization value generation unit 603 (the numbers are the same as those in FIG. 1).
0), decoding is performed according to the algorithm of FIG. 11, and the obtained reproduction value is output to the reproduction value output unit 1060.

As described above, the image encoding / decoding device according to the eleventh embodiment includes the input switching unit 1053 and the output switching unit 1058, and switches input / output during encoding and decoding to perform quantization and quantization. By providing an inverse quantization processing unit 1054 for use in encoding and decoding, the same device performs encoding and decoding so that device resources can be effectively used, and in addition to encoding, Since the data obtained as a result can be reproduced immediately, it is a useful device that can be used for experimental coding and decoding for performing various settings and the like.

In the image encoding / decoding apparatus according to the eleventh embodiment, the quantization / inverse quantization processing section 1054
In addition, a configuration including a shift value generation unit may be provided, and the configuration may be equivalent to the image encoding device and the image decoding device according to the fourth embodiment. In this case, as in the fourth embodiment, it is possible to recover the dynamic range by shifting and reduce the limitation of the dynamic range.

Embodiment 12 FIG. Embodiment 12 of the present invention
In the image encoding / decoding apparatus described in Embodiment 11, the quantization / inverse quantization processing unit shares a circuit for encoding (quantization) processing and decoding (inverse quantization) processing. Is shown.

FIG. 28 is a circuit block diagram showing a configuration of a quantization / dequantization shared circuit of an encoding / decoding device according to the twelfth embodiment. 28, reference numeral 1101 denotes an input pixel value input from the pixel value input unit 1151 of FIG. 27, 1102 denotes an input quantization value input from the quantization value input unit 1152 of FIG. 27, and 1103 denotes a reproduced value of FIG. A reproduction value 1104 output to the output unit 1060 is a quantization value output to the quantization value output unit 1059 in FIG.

In FIG. 28, 1105-1112
Is a switch, 1131-1123 is an adder / subtractor, 112
4 and 1125 are comparators, 1126 and 1127 are delay units. The switch 1105 functions as the input switching unit 1053 in FIG. 27, and the switch 1112 functions as the output switching unit 1058 in FIG. The other components constitute the quantization / dequantization processing unit 1054 in FIG. FIG.
In the circuit (1), at the time of quantization (at the time of encoding), 1105 to 1
The switch 110 is connected to e, and is connected to d at the time of inverse quantization (at the time of decoding). Thereby, at the time of encoding, the same processing as the above-described image encoding algorithm can be realized, and at the time of decoding, the same processing as the above-described image decoding algorithm can be realized.

As described above, in the shared circuit according to the twelfth embodiment, most circuits can be shared by switching between the encoding processing and the decoding processing as shown in FIG. Can be greatly reduced, cost can be reduced, and device resources can be effectively used.

As described above, the image encoding apparatus, the decoding apparatus, and the encoding / decoding apparatus according to the present invention have been described with reference to the numerous embodiments. The present invention is applicable to any image signal, and any method other than the embodiment can be applied to a method of generating a predicted value, the number of bits of an input pixel value and a quantization value, and the like. Also, the configuration described in the block diagram of the embodiment can be realized in various ways including the processing order, and can be realized by software. Furthermore, an image encoding device and a decoding device that combine the above-described techniques are also possible.

[0157]

According to the first aspect of the present invention, there is provided a predictive value generating means for generating a predictive value for an input pixel value from pixels around the input pixel, and a linear quantization having a linear quantization representative point. A linear quantizer generating means for generating a quantizer; and for the linear quantizer, a quantization representative point is added around the predicted value and the quantization width around the predicted value is linearly quantized. And a non-linear quantizer generating means for generating a non-linear quantizer smaller than the non-linear quantizer.
The input pixel value is directly quantized by a simple adder / subtractor and comparator without using an OM table or the like, so that a non-linear quantization process can be realized with a small circuit scale, and each quantized value itself is an absolute value. Since level information is included, there is an effect that error propagation hardly occurs even when a predicted value is incorrect.

According to a second aspect of the present invention, the image encoding apparatus of the first aspect further comprises a shift value generating means for generating a shift value from the predicted value generated by the predicted value generating means. As a result, the shift prediction value is used for the calculation of the quantization process, and in addition to the effect of non-linear quantization and error propagation prevention by a small-scale circuit, the reduction of the dynamic range caused by the non-linear quantization process by adding the quantization representative point Can be recovered by expanding the dynamic range in the shift direction.

According to the image coding apparatus of claim 3, in the apparatus of claim 1, the non-linear quantizer generating means includes:
By adding a quantization representative point for non-linear quantization processing and a quantization representative point for dynamic range recovery, in addition to the effects of nonlinear quantization and error propagation prevention by a small-scale circuit, The reduction of the dynamic range caused by the non-linear quantization processing due to the addition of the quantization representative points can be recovered by adding the quantization representative points to the reduced area.

According to a fourth aspect of the present invention, there is provided a predictive value generating means for generating a predictive value for an input pixel value from pixels around the input pixel, and a division process for the input pixel value with d-bit precision. Linear quantization value generation means for generating a linear quantization value; and a non-linear quantization value generation means for correcting the linear quantization value to generate a non-linear quantization value by a difference value between the input pixel value and the prediction value Means, the input pixel value is directly quantized by a simple adder / subtractor and comparator without using a ROM table or the like, and it is possible to realize a non-linear quantization process with a small circuit scale. In addition, since each quantized value itself contains absolute level information, there is an effect that error propagation hardly occurs even when a predicted value is incorrect.

According to a fifth aspect of the present invention, in the image encoding apparatus of the fourth aspect, a shift value generating means for generating a shift value from the predicted value generated by the predicted value generating means is further provided. As a result, the shift prediction value is used for the calculation of the quantization process, and in addition to the effect of non-linear quantization and error propagation prevention by a small-scale circuit, the reduction of the dynamic range caused by the non-linear quantization process by adding the quantization representative point Can be recovered by expanding the dynamic range in the shift direction.

According to the image coding apparatus of the sixth aspect, in the apparatus of the fourth aspect, when the magnitude of the input pixel value is out of the predetermined range, the predetermined limited quantization value is converted to the nonlinear quantization value. Input value limiting means for outputting to the quantized value generating means, wherein the non-linear quantized value generating means, when the restricted quantized value is input, uses the restricted quantized value as a non-linear quantized value. In addition to the effect of non-linear quantization and error propagation prevention by a small-scale circuit, the reduction of the dynamic range caused by the non-linear quantization processing by adding a quantization representative point, a specific for the reduced region It is possible to recover by outputting the quantized value.

According to a seventh aspect of the present invention, in any one of the first to sixth aspects of the present invention, the offset value is determined based on the remainder of the division of the predicted value, and the linear quantization included in the linear quantizer is performed. At the position of the representative point or with respect to the input pixel value, the linear quantizer generating means or the linear quantizing value generating means has an offset value adding means for adding the above offset value. It is possible to efficiently execute the calculation in the processing.

According to the image coding apparatus of claim 8, in the apparatus of any one of claims 1 to 5, the input value limiting means for limiting the size of the input pixel value and outputting a limited input value is provided. By providing further, the input pixel value is limited to a range that can be processed by linear quantization processing,
In the non-linear quantization processing that prevents error propagation in a small-scale circuit, it is possible to well retain the characteristic of good image quality by predictive coding.

According to a ninth aspect of the present invention, in the apparatus of the second or fifth aspect, the shift input value limiting means for limiting the magnitude of the shift input value and outputting the limited shift input value is further provided. With this feature, the shift input value is limited to a range that can be processed by linear quantization processing, and in non-linear quantization processing that prevents error propagation in small-scale circuits, good image quality can be achieved by predictive coding. In holding the characteristic, more flexible processing can be performed according to the shift value setting.

According to a tenth aspect of the present invention, in the image encoding apparatus of any one of the first to sixth aspects, the quantizing means or the non-linear quantizing value generating means is configured to determine the magnitude of the generated quantizing value. Has a quantization value limiting means for limiting the quantization value obtained as a result of encoding to a certain range, while eliminating the limitation of the dynamic range of the input value, in a small circuit. In the non-linear quantization processing that prevents error propagation, it is possible to well retain the characteristic of good image quality by predictive coding.

According to the image coding apparatus of claim 11, in the apparatus of any of claims 1 to 6, the quantizing means or the non-linear quantizing value generating means corresponds to the case where the input pixel value is minimum. The quantization value is generated so that the quantization value to be obtained becomes the minimum value and the corresponding quantization value becomes larger as the input pixel value becomes larger. Level information is included, and the effect that error propagation hardly occurs even when the predicted value is incorrect is obtained.

According to a twelfth aspect of the present invention, in the image encoding apparatus of any one of the first to eleventh aspects, the quantizing means or the nonlinear quantized value generating means converts the specific n-bit pattern into an error code. Defined as not used for the output quantized value, it is possible to assign a specific pattern to an error code, so that even if an error occurs on the transmission line, By detecting and correcting the error, the effect of the error can be reduced.

The image coding apparatus according to claim 13 is based on claim 8
Or, in the apparatus according to 9, the quantization means or the nonlinear quantization value generation means defines a specific n-bit pattern as an error code and does not use it for the output quantization value. The unused pattern is defined as an error code due to the above restriction, so that good image quality can be obtained and errors can be corrected by the error code during reproduction.

According to the image decoding apparatus of any one of claims 14 to 24, non-linear quantization can be realized without using a ROM table or the like, and the circuit scale can be reduced.

According to the image decoding apparatus of the twenty-first aspect, in the apparatus of any one of the fourteenth to nineteenth aspects, a specific n-bit pattern is defined as an error code that is not used in ordinary encoding, and the input quantum If the pattern of the coded value matches the pattern of the error code defined above, the predicted value is set to the reproduced value, and the error code is detected even when an error occurs in the transmission path. By correcting the error, the influence of the error can be reduced.

According to the image decoding apparatus of claim 22, in the image decoding apparatus of claim 21, when there is a possibility that an error exists in the transmitted input quantization value, the input quantization value Is replaced with the error code defined above, so that error correction can be performed.

According to the image encoding / decoding apparatus of the twenty-third aspect, an input pixel or a pixel around an input quantized value is
A prediction value generation unit that generates a prediction value for an input pixel value or an input quantization value, and a linear processing value generation unit that performs multiplication or division on an input linear processing target value to generate a linear processing value A non-linear processing value generating unit configured to correct the linear processing value to generate a non-linear processing value by a difference value between the determined non-linear processing target value and the linear processing value; The circuit scale can be further reduced by sharing the encoding and decoding circuits, the results of encoding and decoding can be easily confirmed, and experimental coding and experiment for performing various settings are performed. Useful for decryption.

According to a twenty-fourth aspect of the present invention, the image encoding / decoding apparatus according to the twenty-third aspect further comprises a shift value generating means for generating a shift value from the predicted value generated by the predicted value generating means. With this configuration, in addition to the effect of the small-scale circuit, the decrease in the dynamic range can be recovered by expanding the dynamic range in the shift direction.

The image encoding device, the image decoding device, and the image encoding / decoding device of the present invention perform non-linear processing by relatively simple arithmetic processing without using a ROM table or the like. Therefore, it can be realized with a small circuit scale, and can be processed at high speed by software in a general-purpose device such as a personal computer or a workstation.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration of an image encoding device according to Embodiment 1 of the present invention.

FIG. 2 is a diagram for explaining quantization of the device.

FIG. 3 is a block diagram showing a configuration of an image decoding device according to the embodiment.

FIG. 4 is a block diagram illustrating a configuration of an image encoding device according to a second embodiment of the present invention.

FIG. 5 is a diagram illustrating quantization of the device.

FIG. 6 is a block diagram illustrating a configuration of an image decoding device according to the embodiment.

FIG. 7 is a block diagram illustrating a configuration of an image encoding device according to a third embodiment of the present invention.

FIG. 8 is a diagram for explaining quantization of the device.

FIG. 9 is a flowchart showing encoding by the apparatus.

FIG. 10 is a block diagram showing a configuration of an image decoding device according to the embodiment.

FIG. 11 is a flowchart showing decoding by the device.

FIG. 12 is a block diagram illustrating a configuration of an image encoding device according to a fourth embodiment of the present invention.

FIG. 13 is a flowchart showing encoding by the apparatus.

FIG. 14 is a block diagram illustrating a configuration of an image decoding device according to the embodiment.

FIG. 15 is a flowchart showing decoding by the device.

FIG. 16 is a block diagram illustrating a configuration of an image encoding device according to a fifth embodiment of the present invention.

FIG. 17 is a diagram for explaining quantization of the device.

FIG. 18 is a block diagram illustrating a configuration of an image decoding device according to the embodiment.

FIG. 19 is a block diagram illustrating a configuration of an image encoding device according to a sixth embodiment of the present invention.

FIG. 20 is a block diagram illustrating a configuration of an image decoding device according to the embodiment.

FIG. 21 is a diagram for explaining quantization in the embodiment.

FIG. 22 is a block diagram illustrating a configuration of an image encoding device according to a seventh embodiment of the present invention.

FIG. 23 is a block diagram illustrating a configuration of an image encoding device according to Embodiment 8 of the present invention.

FIG. 24 is a flowchart illustrating encoding performed by the image encoding device according to Embodiment 9 of the present invention;

FIG. 25 is a block diagram illustrating a configuration of an image decoding device (decoding portion) according to Embodiment 10 of the present invention.

FIG. 26 is a block diagram illustrating a configuration of an image decoding device (replacement part) according to Embodiment 10 of the present invention;

FIG. 27 is a diagram illustrating an image encoding / coding method according to Embodiment 11 of the present invention;
It is a block diagram which shows the structure of a decoding device.

FIG. 28 is a circuit block diagram showing a configuration of an encoding / decoding shared circuit according to a twelfth embodiment of the present invention.

FIG. 29 is a diagram illustrating linear and nonlinear quantization according to the related art.

[Explanation of symbols]

 101, 1051 Pixel value input unit 102 Linear quantizer generator 103, 801, 901 Non-linear quantizer generator 104 Quantizer 105, 303, 1204 Output unit 106 Predicted value generator 107 Shift value generator 301, 1052 Quantum Quantization value input section 302 inverse quantization section 401 linear quantization value generation section 402, 1002 nonlinear quantization value generation section 601 linear reproduction value generation section 602, 1201 nonlinear quantization reproduction value generation section 603 prediction value linear quantization value generation section 701, 1001 input value limiting unit 702 image encoding unit 703 quantization value control unit 902 error code detection unit 903 prediction error setting unit 904 image decoding unit 1053 input switching unit 1054 quantization / inverse quantization processing unit 1055 linear processing means 1056 prediction value generation means 1057 nonlinear processing means 1058 output switching unit 1059 Quantized value output unit 1060 Reproduction value output unit 1201 Reproduction value input unit 1202 Error code replacement unit 1203 Error detection unit

 ──────────────────────────────────────────────────続 き Continued on the front page (72) Inventor Yuji Fujiwara 1006 Kadoma Kadoma, Osaka Prefecture Matsushita Electric Industrial Co., Ltd.

Claims (24)

[Claims] 1. An image coding apparatus which obtains an n-bit quantized value by encoding from an input pixel value having a dynamic range of d bits and transmits the quantized value, comprising the steps of: Prediction value generation means for generating a prediction value for the value, and a quantization width of 2 ^dn and 2 ^{n in} d-bit precision
A linear quantizer generating means for generating a linear quantizer having a number of linear quantization representative points obtained by subtracting a preset additional upper limit number from the above, and for the linear quantizer, Near the periphery of the value,
By adding the number of quantization representative points equal to or less than the additional upper limit number,
A non-linear quantizer generating means for generating a non-linear quantizer in which the quantization width near the predicted value is smaller than the linear quantizer, and quantizing an input pixel value with the non-linear quantizer to obtain a quantized value An image coding apparatus comprising: a quantization unit for obtaining an image. 2. The image encoding apparatus according to claim 1, further comprising: a shift value generating unit configured to generate a shift value from a predicted value generated by the predicted value generating unit. For the linear quantizer generated by the linear quantizer generating means, a number of quantization representatives equal to or less than the additional upper limit number are provided around a shift predicted value obtained by subtracting the shift value from the predicted value. A non-linear quantizer is generated by adding a point.The quantizing means quantizes a shift input value obtained by subtracting the shift value from an input pixel value by the non-linear quantizer, and An image encoding apparatus for obtaining an encoded value. 3. The image encoding apparatus according to claim 1, wherein the nonlinear quantizer generating unit is configured to:
From the additional upper limit number, the number of quantization representative points equal to or smaller than the number obtained by subtracting the second additional upper limit number set in advance is added, and the quantization width near the predicted value is made smaller than that of the linear quantizer. In addition, the linear quantizer generation unit does not set the linear quantization representative point, and the quantization width is 2 ^dn . An image coding apparatus, wherein a non-linear quantizer is generated by adding a quantization representative point with a quantization width larger than that of the linear quantizer. 4. An image coding apparatus which obtains an n-bit quantized value by encoding from an input pixel value having a dynamic range of d bits and transmits the quantized value, comprising the steps of: Prediction value generation means for generating a prediction value for a value; linear quantization value generation means for generating a linear quantization value by dividing an input pixel value by 2 ^dn with d-bit precision; An image coding apparatus comprising: a non-linear quantization value generation unit that corrects the linear quantization value according to a value difference value to generate a non-linear quantization value. 5. The image encoding apparatus according to claim 4, further comprising: a shift value generating unit configured to generate a shift value from a predicted value generated by the predicted value generating unit; An image coding apparatus, wherein a linearly quantized value is generated by dividing a shift input value obtained by subtracting the shift value from an input pixel value by 2 ^dn with d-bit precision. 6. The image encoding device according to claim 4, wherein the magnitude of the input pixel value is checked, and if the value is within a predetermined range, the input pixel value is sent to the linear quantization value generation means. And outputting the input pixel value when the value is out of the predetermined range, and outputting the predetermined limited quantization value to the non-linear quantization value generation means. Wherein the non-linear quantization value generating means, when the restricted quantization value is input from the input value restricting means, uses the restricted quantization value as a non-linear quantization value. Image encoding device. 7. The image coding apparatus according to claim 1, wherein said linear quantizer generating means or said linear quantized value generating means ^modulates said prediction value by 2 ^dn. Offset value adding means for adding the offset value to the position of the linear quantization representative point of the linear quantizer or to the input pixel value. Image encoding device. 8. The image encoding apparatus according to claim 1, further comprising an input value limiting unit that outputs a limited input value by limiting a size of the input pixel value. An image encoding device, comprising: 9. The image encoding apparatus according to claim 2, further comprising a shift input value limiting unit that outputs a limited shift input value by limiting the size of the shift input value. An image encoding device, characterized in that: 10. The image encoding apparatus according to claim 1, wherein said quantization means or said nonlinear quantization value generation means limits a size of said generated quantization value. An image coding apparatus comprising a quantization value limiting unit. 11. The image coding apparatus according to claim 1, wherein said quantization means or said non-linear quantization value generation means performs quantization corresponding to a case where an input pixel value is minimum. An image coding apparatus, wherein the quantization value is generated such that the value becomes a minimum value and the corresponding quantization value becomes a larger value as the input pixel value increases. 12. The image encoding apparatus according to claim 1, wherein said quantization means or said nonlinear quantization value generation means defines a specific n-bit pattern as an error code. An image encoding apparatus that does not use a quantized value to be output. 13. The image encoding apparatus according to claim 8, wherein said quantization means or said non-linear quantization value generation means defines a specific n-bit pattern as an error code and outputs it. An image coding apparatus characterized in that it is not used for a quantization value. 14. An image decoding apparatus for decoding a n-bit quantized value to obtain a reproduced value having a dynamic range of d bits, comprising: a predictive value for an input quantized value from pixels surrounding the input quantized value; And a prediction value generating means for generating the following formula: In d-bit precision, the quantization width is 2 ^dn and 2 ⁿ
A linear quantizer generating means for generating a linear quantizer having a linear quantization representative point, the number of which is obtained by subtracting a preset additional upper limit number from In the vicinity of,
By adding the number of quantization representative points equal to or less than the additional upper limit number,
A non-linear quantizer generating means for generating a non-linear quantizer in which the quantization width near the predicted value is smaller than the linear quantizer; An image decoding device comprising: an inverse quantization means for obtaining 15. The image decoding apparatus according to claim 14, further comprising: a shift value generating unit configured to generate a shift value from the predicted value, wherein the nonlinear quantizer generating unit is configured to generate a shift value for the linear quantizer. A non-linear quantizer is generated by adding quantization representative points equal to or less than the number in the vicinity of the shift predicted value obtained by subtracting the shift value from the predicted value to generate a nonlinear quantizer. An image decoding apparatus for outputting a pixel value obtained by adding the shift value to the obtained reproduction value. 16. The image decoding apparatus according to claim 14, wherein said nonlinear quantizer generating means is arranged to:
From the additional upper limit number, the number of quantization representative points equal to or smaller than the number obtained by subtracting the second additional upper limit number set in advance is added, and the quantization width near the predicted value is made smaller than that of the linear quantizer. In addition, the linear quantizer generation unit does not set the linear quantization representative point, and the quantization width is 2 ^dn . An image decoding apparatus, wherein a non-linear quantizer is generated by adding a quantization representative point with a larger quantization width than the linear quantizer. 17. An image decoding apparatus which decodes an n-bit quantized value to obtain a reproduced value having a dynamic range of d bits, comprising: predicting a predicted value for an input quantized value from pixels surrounding the input quantized value; A predicted value linearly quantized value generating unit that divides the predicted value by 2 ^dn to generate a predicted value linearly quantized value in d-bit precision; A linear quantization reproduction value generating means for generating a linear quantization reproduction value by multiplying the input quantization value by 2 ^dn, and a prediction value obtained by a difference value between the input quantization value and the prediction value linear quantization value Alternatively, an image decoding apparatus comprising: an inverse quantization means for correcting the linearly quantized reproduction value to generate a reproduction value. 18. The image decoding apparatus according to claim 17, further comprising: a shift value generating unit configured to generate a shift value from the predicted value, wherein the predicted value linearly quantized value generating unit is configured to calculate the shift value from the predicted value. The predicted shift value obtained by subtracting the shift value is divided by 2 ^dn to generate a linearly quantized predicted value. The shift value is added to the reproduction value generated by the inverse quantization means. An image decoding device for outputting a pixel value. 19. The image decoding apparatus according to claim 17, wherein said linear quantization reproduction value generation means generates a predetermined linear quantization reproduction value when said input quantization value is a predetermined value. An image decoding apparatus characterized in that: 20. The image decoding apparatus according to claim 14, wherein said linear quantizer generating means or said linear quantized reproduction value generating means divides said prediction value by 2 ^dn . An offset value adding means for determining an offset value based on a remainder and adding the offset value to the position of the linear quantization representative point of the linear quantizer or to the linear quantization reproduction value. An image encoding device, characterized in that: 21. The image decoding apparatus according to claim 14, wherein a specific n-bit pattern is defined as an error code that is not used in normal encoding. An image decoding apparatus characterized in that, when a pattern having a value matches an error code pattern defined above, the predicted value is used as the reproduction value. 22. The image decoding device according to claim 21, wherein when there is a possibility that an error exists in the input quantization value to be transmitted, the input quantization value is replaced with the defined error code. An image decoding apparatus characterized in that: 23. An n-bit quantized value is obtained by encoding from an input pixel value having a d-bit dynamic range and transmitted, and the n-bit quantized value is decoded to obtain a dynamic range. An image encoding / decoding device for obtaining a d-bit reproduction value, comprising: a prediction value generation unit configured to generate a prediction value for an input pixel value or an input quantization value from an input pixel or a pixel around the input quantization value. A linear processing value generating means for performing a multiplication or division by 2 ^dn on the input linear processing target value with d-bit precision to generate a linear processing value; An image encoding / decoding apparatus, comprising: a non-linear processing value generating unit that corrects the linear processing value based on a difference value between the processing value and a non-linear processing value. 24. The image encoding / decoding apparatus according to claim 23, further comprising: a shift value generating unit configured to generate a shift value from the predicted value.
Decryption device.