JPH0937242A - Variable bit rate image signal coder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide the variable bit rate image signal decoder in which image quality of decoded image is made constant. SOLUTION: A characteristic quantity calculation section 122 calculates K kinds of characteristic quantities, and a frame unit statistic variable calculation section 123 calculates a statistic variable 124 in a frame whose characteristic variable is measured. Based on the statistic variable 124 of each characteristic variable of the frame to be calculated, a frame group statistic variable calculation section 125 calculates a statistic variable of each characteristic variable in a received frame group. An image quality deterioration estimate section 127 estimates an image deterioration degree 128 of a coding object frame based on a mean value 126 of the statistic variable of each characteristic variable in the frame group. A quantization step calculation section 112 calculates a quantization step 119 based on the image deterioration degree 128, a frame group maximum incidence information variable 115, a frame group average incidence information variable 116, a maximum rate 17, and an average rate 118.

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention comprises a generated information amount measuring unit and an encoding unit, which divides a digital image signal inputted by the generated information amount measuring unit into small blocks and encodes them in a fixed quantization step. The present invention relates to a variable bit rate image signal coding device that measures the amount of generated information, calculates a quantization step in a coding unit from the measurement result, and performs coding.

[0002]

2. Description of the Related Art Variable bit rate coding control has less restrictions on the amount of generated information as compared with fixed bit rate and allows flexible bit allocation, so that the image quality is almost the same for image signals having various characteristics. Can be encoded in.
In the variable bit rate coding control, the average rate, which is the amount of generated information within a fixed time T _a , and the maximum rate, which is the amount of generated information within a time T _m shorter than T _a , are used as parameters. In variable bit rate coding control, information allocation is optimized using the above control parameters, so the frame to be coded is analyzed in advance to measure the amount of generated information, and the quantization step is determined from the measurement result. There is a two-pass coding control that performs coding after that.

FIG. 7 is for explaining an example of a conventional two-pass coding control method. In this example, it is assumed that the discrete cosine transform is used as the image coding method.
A plurality of frames to be measured in advance is called a frame group, and the number of frames in the frame group is F. The control parameters are a maximum rate B _max indicating an upper limit of the generated information amount in a frame unit and an average upper limit of the average of all generated information amounts of the frame group per frame as an average rate B _ave .

First, in the generated information amount measuring unit, the input terminal 1
The generated information amount of the image signal input from 01 is measured. The image signal input to the generated information amount measuring unit is divided into N × N small blocks by the small block dividing unit 102. The discrete cosine transform unit 103B performs discrete cosine transform on the divided small block image signal. Further, a preset quantization step Q _p (10
5), the discrete cosine transform coefficient is calculated by the quantization unit 104.
Is quantized by, and the variable length coding unit 106 performs variable length coding, and the generated information amount calculation unit 107 measures the generated information amount 113 for each frame. Generated information amount statistic calculation unit 11
The frame group maximum generated information amount G _max (115) and the frame group average generated information amount G _ave (116) are calculated from the frame-based generated information amount 113 measured in 4. The frame group input to the generated information amount measuring unit is stored in the frame memory 108 until the measurement of the generated information amount of all frames in the frame group is completed.

Next, after the measurement of the generated information amount is finished, the image signal stored in the frame memory 108 is coded by the coding unit. Here, the frame number in the frame group of the image to be encoded is i (i = 1, 2, ..., F). The image signal input to the encoding unit is divided by the small block dividing unit 109 into N × N small blocks, and the discrete cosine transform unit 1 is applied to the divided small block image signal.
Discrete cosine transform is performed at 10B. On the other hand, the frame group maximum generated information amount G measured by the generated information amount measuring unit
_max (115) and frame group average generated information amount G _ave
(116) and the maximum rate B _max (117) and the average rate B _ave (11) given as control parameters.
From 8), the quantization step calculation unit 112 calculates the reference quantization step Q _i (119). The calculation method is as follows, and it is assumed that the quantization step Q and the generated information amount G have a relationship of G = a / Q. Here, a is a constant.

For the average rate B _ave (118), the average generated information amount at the time of encoding is B _ave (115) from the frame group average generated information amount G _ave (116) and the quantization step Q _p (105) used for measurement. ), The quantization step Q _i (119) is calculated. Q _i = Q _p · G _avev / B _ave However, when quantized in the calculated quantization step Q _i (119), the amount of generated information is increased for a frame exceeding the maximum rate B _max (115). A quantization step Q _i (119) that does not exceed G _max is newly set.

Q _i = Q _p · G _max / B _{max The} discrete cosine transform coefficient is determined by the determined quantization step Q.
_{With i} (119), the quantization unit 111 quantizes and the variable length coding unit 120 performs variable length coding.

According to such a method, the quantization steps can be made substantially uniform, and as a result, the image quality difference between decoded images becomes small.

[0009]

In the conventional method, the quantization step is made constant in order to make the image quality of the decoded image constant, but the actual image quality of the decoded image is not always constant. This is due to human visual characteristics, and the image quality looks different depending on the nature of the image signal.

For example, as shown in FIG. 6, in a frame in which the luminance signal changes little and in a frame in which the luminance signal changes a lot, even if the quantization step is the same, the degree of image quality deterioration appears different.
This is known as the masking effect.

In the conventional method, since the quantization step is controlled to be constant in order to keep the image quality deterioration constant, when there are image signals having different characteristics, the quantization is performed at the same quantization step. Even the image may look different in image quality. This indicates that bit allocation is not performed so as to keep the image quality constant. That is, extra information is assigned to one frame and less information is assigned to another frame.

As described above, in the conventional method, since the image quality is kept constant by keeping the quantization step constant regardless of the visual characteristic, there is a problem that proper bit allocation is not performed.

It is an object of the present invention to provide a variable bit rate image signal coding device capable of keeping the image quality of a decoded image constant.

[0014]

A variable bit rate image signal encoding apparatus according to the present invention calculates a feature amount of one kind or a plurality of kinds for each small block obtained by dividing an image signal of each input frame. Calculating means, and a first statistical calculating means for calculating a statistical amount of the characteristic amount in a certain coding unit,
Second calculation of statistics for each feature over multiple frames
Statistic calculating means, image quality deterioration estimating means for estimating the degree of deterioration of the image quality of the encoding target frame from the statistic of the encoding unit of the feature and the statistic of each feature over a plurality of frames, and the occurrence information. It has a quantization step calculation means for calculating and setting a quantization step for each coding unit from the amount and the degree of deterioration of the image quality.

The quantization step calculation means calculates the degree of deterioration of the image quality (D _i ), the maximum amount of generated information (G _max ) and the average amount of generated frame group (G _ave ) which are the generated information amounts. And the maximum rate (B _max ) that is the upper limit of the amount of generated information in frame units, which is a control parameter
And the average rate (B _ave ) that is the upper limit of the average per frame of all the generated information amount of the frame group and the quantization step (Q _p ) used for measuring the generated information amount, the average generated information amount at the time of encoding. Determines a quantization step Q _i ′ = Q _p · G _ave / B _ave that does not exceed the average rate (B _ave ), and then determines a quantization step Q _i = Q _i ′ · D _i. When quantized in the quantization step, the maximum rate B _max
For the frame exceeding the amount of information generated does not exceed the maximum rate B _max quantization step _{Q i = Q p · G max} /
B _max is newly set.

The feature amounts are the intra-frame luminance signal variance and the magnitude of the motion vector.

The image quality deterioration estimating means estimates the degree of deterioration of the image quality of the encoding target frame from the average value of both characteristic amounts of the encoding target frame and the average value of both characteristic amounts of the frame group.

The quantizing step calculating means includes the frame group maximum generated information amount (G _max ) and the frame group average generated information amount (G _ave ) which are the generated information amounts, and the generated information in frame units which is a control parameter. The maximum rate (B _max ) which is the upper limit of the amount and the average rate (B which is the upper limit of the average per frame of all the generated information amount of the frame group
_{av e} ) and the quantization step (Q _p ) used when measuring the generated information amount, the average quantization step Q _all = Q _p at which the average generated information amount during encoding does not exceed the average rate (B _ave ).
· Defining a G _ave / B _ave, when the quantized by the calculated average quantization step Q _ave, with respect to the frame exceeding the maximum rate B _max, the average quantization step quantity occurrence information does not exceed B _max It has an average quantization step calculation means for newly setting Q _ave = G _p · G _max / B _max, and the quantum of the small block to be encoded is determined from the degree of deterioration D of the image quality and the average quantization step Q _all. Conversion step Q = Q _all
・ Determine D.

The image quality deterioration estimating means estimates the degree of image quality deterioration of the encoding target small block from the characteristic amount of the encoding target small block and the average value of all the small blocks in the frame group of the characteristic amount.

[0020]

BEST MODE FOR CARRYING OUT THE INVENTION The image signal coding apparatus of the present invention comprises a generated information amount measuring section and a coding section. The generation information measuring unit checks the amount of information generated when the quantization at a predetermined image signal of each frame input quantization step Q _P, and determines the quantization control that reflects the visual characteristics Since it is performed in the specified encoding unit, the characteristic amount in the encoding unit and the statistical amount of the characteristic amount in the frame group are calculated from one or a plurality of characteristic amounts for each small block. The encoding unit estimates the degree of deterioration of the image of the encoding target frame from the characteristic amount of the encoding unit to be encoded and the statistical amount of the characteristic amount in the frame group, and determines the degree of image deterioration and the occurrence information. The quantization step of the coding unit to be coded is set from the amount and the control parameter, and coding is performed.

According to such a method, the quantization step can be set according to the visual characteristic, and as a result, the image quality difference between the decoded images becomes small.

[0022]

Next, embodiments of the present invention will be described with reference to the drawings.

FIG. 1 is a block diagram of an image signal coding apparatus according to a first embodiment of the present invention.

Let F be the number of frames in the frame group to be measured, and K be the number of feature quantities to be used. As the control parameters, the upper limit of the generated information amount in frame units is the maximum rate B _max , and the upper limit of the average of all generated information amounts of the frame group per frame is the average rate B _ave .

First, in the generation information measuring unit, the input unit 10
The generated information amount of the image signal input from 1 is measured.
The image signal input to the generated information amount measuring unit is divided into N × N small blocks by the small block dividing unit 102. The orthogonal transformation unit 10 applies to the divided small block image signal.
Orthogonal transformation is performed at 3A. Further, the orthogonal transform coefficient is quantized by the quantization unit 104 using a preset quantization step Q _p (105), and the variable length coding unit 1
Variable length coding is performed at 06, and the generated information amount calculation unit 107 measures the generated information amount. On the other hand, the feature amount calculation unit 122 (feature amount calculation unit) calculates K types of feature amounts, and the frame unit statistic amount calculation unit 123 (first statistic amount calculation unit) calculates each feature amount in the measurement frame. Statistics fs _{i, k} (124)
Is calculated. Here, i is a frame number (i = 1, 2, ..., F) in the frame group, and k is a feature number (k =
1, 2, ..., K). The statistic f of each feature amount of the frame calculated by the frame unit statistic calculation unit 123
Based on s _{i, k} (124), the frame group statistic calculation unit 12
5A (second statistic calculation means) calculates the statistic fcs _k (126) of each feature in the input frame group. The frame group input to the generated information amount measuring unit is stored in the frame memory 108 until the measurement of the generated information amount of all frames in the frame group is completed.

Next, after the measurement of the generated information amount, the image signal stored in the frame memory 108 is encoded by the encoding unit. The image signal input to the encoding unit is divided into N × N small blocks by the small block division unit 109, and the orthogonal transformation unit 110A performs orthogonal transformation on the divided small block image signals. On the other hand, in the image quality deterioration estimation unit 127 (image quality deterioration estimation means), the statistical value fs _{i, k} (124) of each feature amount of the encoding target frame and the average value of the statistical amount of each special feature amount in the frame group fcs _k (126)
From the degree of image quality deterioration of the encoding target frame D _i (12
8A) is estimated. D _i is set to be smaller than 1 in a portion where the image quality deterioration is easily noticeable and larger than 1 in a portion where the image quality deterioration is not noticeable.

[0027]

[Equation 1] Next, the degree of image quality deterioration of the encoding target frame D _i (12
8A), frame group maximum generated information amount G _max (115) and frame group average generated information amount G _ave (116) measured by the generated information amount measuring unit, and maximum rate B _max (117) given as a control parameter. ) And the average rate B _ave (118), the quantization step calculation unit 112
In the (quantization step calculation means), the quantization step Q _i
(119) is calculated. The calculation method is as follows.

Average rate B_ave About (118)
Lame group average generated information amount G_ave Used for (116) and measurement
Quantization step Q_p From, the average amount of information generated during encoding
Is B _ave Quantization step Q not exceeding (118)_i ’
Establish.

Q _i '= Q _p · G _ave / B _ave And the degree of image quality deterioration of the encoding target frame D _i (1
28A), the quantization step Q _i (119) is determined as follows.

Q _i = Q _i ′ · D _i The degree of image quality deterioration D _i (128 A) is smaller than 1 in the part where the image quality deterioration is noticeable, so that it is quantized more finely and the image quality deterioration occurs. Since the inconspicuous part has a value larger than 1, it is quantized more coarsely.

However, the calculated quantization step Q_i 
Maximum rate B when quantized in (119)_max (11
For frames exceeding 7), the maximum amount of generated information is
Card B _max Quantization step Q not exceeding (117)_i 
(119) is newly set.

Q _i = G _p G _{max /} B _{max The} orthogonal transform coefficient is obtained by the obtained quantization step Q _i (11
9), the quantization unit 111 quantizes the data, and the variable length coding unit 120 performs variable length coding.

According to this method, the quantization step can be set according to the visual characteristic, and as a result, the image quality difference between the decoded images becomes small.

FIG. 2 is a block diagram of an image signal coding apparatus according to the second embodiment of the present invention. The same reference numerals as those in FIG. 1 indicate the same components.

In this embodiment, two types of feature quantities (intra-frame luminance signal dispersion and motion vector size) are used. Further, an average value is used as a statistic amount for each frame and each frame group. It is assumed that the cosine transform is used as the image coding method. The number of frames in the frame group to be measured is 15. Also, as the control parameters,
Maximum rate B for the upper limit of the amount of information generated in frame units
_{Let max be} the average rate B _ave, which is the upper limit of the average of all the generated information amounts of the frame group per frame.

Only the differences from the first embodiment will be described below.

For the image signal of the small block divided by the small block dividing unit 102, the discrete cosine transform unit 103
The discrete cosine transform is performed at B.

On the other hand, the luminance variance calculating unit 129A and the motion amount calculating unit 130A (feature amount calculating means) calculate the variance of the luminance value and the magnitude of the motion vector for each small block. The variance var of the luminance value of the small block is calculated by the following formula.

[0039]

[Equation 2] The motion amount mv is calculated by comparison with the immediately preceding frame in the frame memory 108, and is calculated by the following equation.

[0040]

(Equation 3)Variance of the calculated small block brightness and motion vector
The size is determined by the frame unit average value calculation unit 131 (first series).
Luminance in the measurement frame for both feature quantities
Average value of value variance var_ave_i (132) and flatness
Average value mv_ave_i Calculate (133). Where i
Are frame numbers (i = 1, 2, ...) In the frame group.
., 15). In the frame unit average value calculation unit 131
Average value var_av of luminance value variance and motion amount calculated by
e_i (132) and mv_ave _i Based on (133)
Frame group average value calculator 134A (second statistic calculator)
), The flatness of both feature quantities in the input frame group.
Average value

[0041]

[Outside 1] Is calculated.

Further, the frame group input to the generated information amount measuring unit is stored in the frame memory 108 until the measurement of the generated information amount of 15 frames to be measured is completed.

Next, after the generation information has been measured, it is encoded by the encoder. The image signal input to the encoding unit is subjected to discrete cosine transform in the discrete cosine transform unit 110B for the image signal of the small block divided by the small block divider 109. On the other hand, in the image quality deterioration estimation unit 127B (image quality deterioration estimation means), the average values var_ave _i (132) and mv_ave of both feature quantities of the encoding target frame.
_i (133) and the average value of both feature quantities of the frame group

[0044]

[Outside 2] From the degree of image quality deterioration of the encoding target frame D _i (12
8B) is estimated. For luminance value dispersion, the higher the dispersion, the less noticeable the image quality deterioration is, and for the amount of motion,
The larger the amount of movement, the less noticeable the image quality deterioration. Therefore, the degree of image quality deterioration D _i (128 B) is calculated from the following equation.

[0045]

(Equation 4) The degree of image quality deterioration D _i (128B) is 1 in the average part, but is 1/4 in the part where the image quality deterioration is most noticeable.
And becomes 4 at the part where the image quality deterioration is most unnoticeable. Next, the degree of image quality deterioration D _{i of the} encoding target frame
(128B), the frame group maximum generated information amount G _max (115) and the frame group average generated information amount G _ave (116) measured by the generated information amount measuring unit, and the maximum rate B _max (given as a control parameter). 117) and the average rate B _ave (118), the reference quantization step calculation unit 112B (quantization step calculation means) calculates the reference quantization step Q _i (119) by the same calculation method as in the first embodiment. Calculate with.

In the second embodiment described above, the number of feature amounts was set to 2, and the feature amount was the luminance dispersion and the magnitude of the motion vector, but the feature amount and the type of feature amount are not limited to this. Furthermore, although the statistic amount in frame units and frame groups is the average value, other statistic amounts such as variance may be used. Furthermore, the method of calculating the degree of image quality deterioration and the method of calculating the quantization control parameter can be arbitrarily determined. Also, with the interframe coding method,
It can be used as in the present embodiment.

FIG. 3 is a block diagram of an image signal coding apparatus according to the third embodiment of the present invention. The same reference numerals as those in FIGS. 1 and 2 denote the same components.

Let F be the number of frames in the frame group to be measured, B be the number of small blocks per frame, and K be the number of feature quantities to be used. FIG. 4 shows the relationship between the frame group and the small blocks. As the control parameter, the upper limit of the generated information amount in the frame unit is the maximum rate B _max , and the upper limit of the average of the generated information amount in the frame group in the frame unit is the average rate B _ave .

Only the parts different from the first and second embodiments will be described below.

The characteristic amount calculation section 137 (characteristic amount calculating means) calculates K types of characteristic amounts fs _{j, k} (138) for each small block. j is a small block number (j = 1, 2, ...
., MF), k is the feature number (k = 1, 2, ...,
K) is represented. Based on each feature amount fs _{j, k} (138) of the frame calculated by the feature amount calculating unit 137, the frame group input by the frame group statistic calculating unit 125B (second statistic calculating unit) Statistic fcs _k of each feature in
(126) is calculated.

In the average quantization step calculation unit 141 (quantization step calculation means), the frame group maximum generation information amount G _max (115) and the frame group average generation information amount G _ave measured by the generation information amount measurement unit are calculated. (116), the maximum rate B _max (11 given as the control parameter)
7) and the average rate B _ave (118), an average quantization step Q _all (142) which is a quantization step for quantizing the entire frame group is calculated. As a calculation method,
It is as follows.

Average rate B_ave About (118)
Lame group average generated information amount G_ave Used for (116) and measurement
Quantization step Q_p From, the average amount of information generated during encoding
Is B _ave Average quantization step Q not exceeding (115)
_all (142) is defined.

Q _all = Q _p · G _ave / B _ave However, the calculated average quantization step Q _all (14
When the quantization is performed in 2), the generated information amount is B _max (11) for the frame exceeding the maximum rate B _max (117).
The average quantization step Q _all (119) that does not exceed 7) is newly set. Q _all = Q _p · G _max / B _max On the other hand, in the feature amount calculation unit 139 (first statistic amount calculation means), K types of feature amounts fs _k (1 of the small block to be encoded) are calculated.
40) is calculated. In the image quality deterioration estimation unit 128 (image quality deterioration estimation means), each feature amount fs _{k of the} small block to be encoded.
(140) and the statistical amount fcs of each feature amount in the frame group
_{From k} (126), the degree D (128c) of image quality deterioration of the small block to be encoded is estimated. The degree of image deterioration D is
It is set so that the part where the image quality deterioration is noticeable is smaller than 1 and the part where the image quality deterioration is less noticeable is larger than 1.

[0054]

(Equation 5) In the quantization step calculation unit 112C (quantization step calculation means), the degree of image quality degradation D (1
28C) and the average quantization step Q _all (142),
The quantization step Q (119) of the small block to be encoded is determined as follows.

Q = Q _all · D The degree of image quality deterioration D (128C) is smaller than 1 in the part where the image quality deterioration is conspicuous, so it is quantized more finely, and the part where the image quality deterioration is not noticeable. Becomes a value larger than 1, so that it is quantized more coarsely.

According to such a method, the quantization step can be set according to the visual characteristic, and as a result, the image quality difference between the decoded images becomes small.

FIG. 5 is a block diagram of an image signal coding apparatus according to the fourth embodiment of the present invention.

In this embodiment, two types of feature quantities (intra-frame luminance signal dispersion and motion vector size) are used. The average value is used as the statistic amount for each frame and each frame group. It is assumed that the cosine transform is used as the image coding method. Let F be the number of frames to be measured and B be the number of small blocks in the frame. As the control parameter, the upper limit of the generated information amount in the frame unit is the maximum rate B _max , and the upper limit of the average of the generated information amount in the frame group in the frame unit is the average rate B _ave . Only parts different from the third embodiment will be described below.

Brightness variance calculator 129A and motion amount calculator 1
30A (feature amount calculation means), the variance var _j (144A) of the luminance value and the motion amount mv _{j for} each small block.
(143A) is calculated. j is a small block number (j =
1, 2, ..., MF). The variance var _j (144A) of the luminance value of the small block is calculated by the following equation.

[0060]

(Equation 6) The motion amount mv _j (143A) is a motion vector of a small block obtained by comparison with the immediately preceding frame in the frame memory 108.

[0061]

[Outside 3] Is calculated and calculated from the following formula.

[0062]

(Equation 7) Variance of brightness values calculated for each small block var _j (14
4A) and the movement amount mv _j (143A), the frame group average value calculation unit 134B (second statistic calculation means)
Average value of both feature values of all small blocks in the input frame group

[0063]

[Outside 4] Is calculated by the following formula.

[0064]

(Equation 8) The image signal input to the encoding unit is stored in the frame memory 10
8A and is divided into N × N small blocks by the small block dividing unit 109, and the discrete cosine transform unit 110B performs discrete cosine transform on the divided small block image signal. On the other hand, the brightness variance calculation unit 129
B and the motion amount calculation unit 130B (first statistic amount calculation means), respectively, the variance var of the luminance values of the encoding target block.
(144B) and the movement amount mv (143B) are calculated.
Subsequently, the image quality deterioration estimation unit 127D (image deterioration estimation means)
, Both feature quantities var (144
B) and mv (143B), and the average value of all small blocks in the frame group of both feature quantities

[0065]

[Outside 5] From the degree of deterioration D (12
8D) is estimated. For luminance value dispersion, the higher the fraction, the less noticeable the image quality deterioration is, and for the amount of movement,
The larger the amount of movement, the less noticeable the image quality deterioration. Therefore, the degree D (128D) of image quality deterioration is calculated by the following equation.

[0066]

[Equation 9] The degree of image quality deterioration D (128D) is 1 in the average part.
However, the part where the image quality deterioration is most noticeable becomes 1/4, and the part where the image quality deterioration is least noticeable becomes 4.
Next, the frame group maximum generated information amount G _max (115) and the frame group average generated information amount G _ave (116) measured by the generation information measuring unit, and the maximum rate B _max (117) given as a control parameter. And average rate B
_{From ave} (118), the average quantization step calculation unit 141
In (quantization step calculation means), an average quantization step Q which is a quantization step for quantizing the entire frame group.
_All (119) is calculated by the same calculation method as in the third embodiment.

Then, the quantization step calculation unit 112C
At (quantization step calculating means), the degree of image degradation of the encoding target small block D from (128D) and the average quantization step Q _{al l} (142), the encoding target small block of the quantization step Q (119) Are determined in the same manner as in the third embodiment.

With such a method, it is possible to estimate the degree of image quality deterioration of a small block in a spatiotemporal image called a frame group, and to determine the quantization step based on the degree of image quality deterioration.

In the fourth embodiment described above, the number of feature amounts was set to 2, and the feature amount was set to the luminance dispersion and the magnitude of the motion vector. However, the number of feature amounts and the types of feature amounts are not limited to this. . Further, although the statistic amount in the frame group is the average value, other statistic amounts such as variance may be used.
Furthermore, the method of calculating the degree of image quality deterioration and the method of calculating the quantization control parameter can be arbitrarily determined. Also, other coding methods such as the interframe coding method can be used as in the present embodiment.

[0070]

As described above, according to the present invention, the deterioration of the image quality due to the coding of the small blocks in the spatiotemporal image is estimated from the characteristics of the image signal, and the quantization step is set for each small block. Therefore, in the area where deterioration is apt to stand out,
Degradation can be made inconspicuous by performing fine quantization, and the amount of information can be reduced by performing coarse quantization in a region where degradation is inconspicuous. As a result, it is effective as a means for allocating an appropriate amount of information when the image signal is encoded, and visual deterioration can be reduced.

[Brief description of drawings]

FIG. 1 is a configuration diagram of a variable bit rate image signal encoding apparatus according to a first embodiment of the present invention.

FIG. 2 is a configuration diagram of a variable bit rate image signal encoding apparatus according to a second embodiment of the present invention.

FIG. 3 is a configuration diagram of a variable bit rate image signal encoding apparatus according to a third embodiment of the present invention.

FIG. 4 is a diagram showing a relationship between a frame group and small blocks.

FIG. 5 is a configuration diagram of a variable bit rate image signal encoding device according to a fourth embodiment of the present invention.

FIG. 6 is a diagram showing an example of images with different luminance changes.

FIG. 7 is a configuration diagram of a conventional example of a variable bit rate image signal encoding device.

[Explanation of symbols]

101 Input Terminal 102 Small Block Dividing Unit 103A, 110A Orthogonal Transforming Unit 103B, 110B Discrete Cosine Transforming Unit 104 Quantizing Unit 105 Quantization Step 106 Variable Length Encoding Unit 107 Generated Information Amount Computing Unit 108, 108A Frame Memory 109 Small Block Dividing Part 111 Quantizer 112A, 112B, 112C, 112D Quantization step calculator 113 Frame-by-frame generated information amount 114 Generated information amount statistic calculator 115 Frame group maximum generated information amount 116 Frame group average generated information amount 117 Maximum rate 118 Average rate 119 Quantization step 120 Variable length coding unit 121 Output terminal 122 Each feature amount calculation unit 123 Frame unit statistic calculation unit 124 Frame unit statistic amount 125A, 125B frame group statistic calculation unit 12 of each feature amount Frame group statistics of each feature amount 127A, 127B, 127C, 127D Image quality deterioration estimation unit 128A, 128B, 128C, 128D Degree of image quality deterioration 129A, 129B Luminance variance calculation unit 130A, 130B Motion amount calculation unit 131 Frame unit average value calculation Part 132 Luminance variance frame unit average value 133 Motion amount frame unit average value 134A, 134B Frame group average value calculation unit 135A, 135B Luminance variance frame group average value 136A, 136B Motion amount frame group average value 141 Average quantization Step calculation unit 142 Average quantization step 143A, 143B Motion amount 144A, 144B Dispersion of luminance value

Claims (6)

[Claims] 1. A digital image signal of each input frame is divided into small blocks, coded at a fixed quantization step, the amount of generated information is measured, and the quantization step is calculated from the measurement result to obtain the digital signal. In a variable bit rate image signal encoding device for encoding an image signal, a feature amount calculation means for calculating one or more types of feature amounts for each small block obtained by dividing an image signal of each input frame, and a certain code A first statistic calculating unit that calculates a statistic of the feature amount in a unit of encoding; a second statistic calculating unit that calculates a statistic of the feature amount over a plurality of frames; An image quality deterioration estimating means for estimating the degree of deterioration of the image quality of the encoding target frame from the statistical amount and the statistical amount of each characteristic amount over a plurality of frames, and the generated information amount and the degree of deterioration of the image quality. Calculating a quantization step of the encoding unit from the variable bit-rate image signal encoding apparatus characterized by having a quantization step calculating unit configured to set. 2. The quantization step calculation means is configured to
Degradation of quality (D _i) And the amount of generated information
Maximum amount of generated information (G_max) And frame group average
Amount of information generated (G_ave) And the control parameter
Maximum rate, which is the upper limit of the amount of generated information
(B_max) And the frame of the total amount of generated information of the frame group
Average rate (B_ave) And before
The quantization step (Q_p)
From the average rate (B
_ave) Quantization step Q not exceeding_i ’= Q_p・ G_ave/
B_ave, Then the quantization step Q = Q_i ’・ D_i To
When quantized by the determined quantization step,
The maximum rate B_max If the frame exceeds
The amount of raw information is the maximum rate B_maxQuantization step that does not exceed
Up Q_i= Q_p・ G_max/ B_maxClaims to newly set
1. The encoding device according to 1. 3. The encoding device according to claim 1, wherein the feature amount is an intra-frame luminance signal variance and a magnitude of a motion vector. 4. The image quality deterioration estimating means estimates the degree of deterioration of the image quality of the encoding target frame from the average value of both characteristic amounts of the encoding target frame and the average value of both characteristic amounts of the frame group. Item 3. The encoding device according to item 3. 5. The quantization step calculation means calculates the amount of generated information in the frame group maximum generated information amount (G _max ), the frame group average generated information amount (G _ave ), and the control parameter in units of frames. The maximum rate (B _max ) which is the upper limit of the generated information amount and the average rate (B which is the upper limit of the average per frame of all the generated information amount of the frame group)
_ave ) and the quantization step (Q _p ) used when measuring the generated information amount, the average quantization step Q _all = where the average generated information amount during encoding does not exceed the average rate (B _ave ).
When Q _p · G _ave / B _ave is determined and quantized by the calculated average quantization step (Q _all ), the maximum rate B _{max is obtained.}
For a frame exceeding _1.0 , the average quantization step Q _all = G _p · G _max / B in which the amount of generated information does not exceed B _max.
An average quantization step calculation means for newly setting _max is provided, and the quantization step Q of the small block to be encoded is calculated from the degree of deterioration D of the image quality and the average quantization step Q _all.
The coding device according to any one of claims 1, 3, and 4, wherein = Q _all · D is obtained. 6. The image quality deterioration estimating means estimates the degree of deterioration of the image quality of the encoding target small block from the feature amount of the encoding target small block and the average value of all the small blocks in the frame group of the feature amount. The encoding device according to claim 5.