JP2001231017A - Absolute image quality evaluation method using electronic watermark, coding method for it, and coder - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an absolute image quality evaluation method that can evaluate the image quality of a decoded image by using electronic watermark information even without an original image to be referenced. SOLUTION: A switch section 10 extracts at random a block Bi from an input image signal 1 by the number of DCT coefficients of one block. A DCT section 13 calculates an i-th DCT coefficient FBi[i] and a bit stream processing section 14 converts the DCT coefficient Fbi[i] into a bit stream dk (k=1-11). A watermark imbed section 15 imbeds a low-order 8-bit of the bit stream dk to a high frequency region of a quantization coefficient matrix of an adjacent block as watermark information together with a code bit. A receiver side detects the watermark information by the number of the DCT coefficients of one block, recovers each virtual block on the basis of the detected watermark information and the DCT correction corresponding to the watermark information obtained from a decoded image to estimate the SNR of the decoded image.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an absolute image quality evaluation method using an electronic watermark, an encoding method therefor, and an encoding apparatus. The present invention relates to an absolute image quality evaluation method using a digital watermark that estimates a SNR of a decoded image by embedding the watermark as a watermark and comparing the coefficient with the coefficient of the decoded image, an encoding method therefor, and an encoding device.

[0002]

2. Description of the Related Art Methods for evaluating the coding quality of digital moving pictures are roughly classified into subjective evaluation and objective evaluation. Subjective evaluation is often used because it can reflect human visual characteristics, but it requires a lot of evaluators and statistical processing, needs to test many evaluation images, There is a problem that also takes.

[0003] On the other hand, objective evaluation is inferior to subjective evaluation in that it reflects human visual characteristics, but the evaluation result is uniquely given as a numerical value by relatively easy calculation, so that it is easy to understand intuitively. There are advantages.

[0004]

However, the conventional subjective evaluation and objective evaluation both evaluate the decoded image by comparing it with the original image. However, since there is no original picture, only a decoded picture is provided, there is no original picture to be referred to, and there is a problem that evaluation by relative evaluation cannot be performed.

An object of the present invention is to solve the above-mentioned problems of the prior art, and to provide an absolute image quality evaluation method capable of evaluating the image quality of a decoded image using digital watermark information even if there is no original image to be referred to. An object of the present invention is to provide an encoding method and an encoding device for that.

[0006]

In order to achieve the above object, the present invention extracts a plurality of blocks per frame from an input image, extracts DCT coefficients from the plurality of blocks, and extracts the extracted DCT coefficients. The first feature is that bit strings are formed, and the obtained bit strings are embedded in quantization matrices of adjacent blocks of the plurality of blocks and transmitted. According to this feature, since the bit string of the DCT coefficient as the watermark information is embedded in the adjacent block, encoding can be performed without affecting the DCT coefficient used for image quality evaluation. Therefore, the estimation accuracy of the image quality evaluation can be improved.

Further, according to the present invention, a DCT is provided.
Receiving an encoded image signal including a block from which a coefficient is extracted and an adjacent block in which a bit string is embedded, and reconstructing a DCT coefficient of the reproduced block and a bit string detected from the adjacent block and complemented with upper bits. The second feature is that the SNR of the decoded image is estimated based on the DCT coefficient reproduced based on the DCT coefficient. According to this feature, with a simple configuration using a general-purpose decoding device, the S
NR can be determined.

The present invention also provides a block extracting means for extracting a plurality of blocks from an input image signal, an orthogonal transform means for orthogonally transforming the blocks extracted by the block extracting means, and an orthogonal transform means. A third feature is that a bit string generating means for converting the orthogonal transform coefficient into a bit string, and a watermark information embedding processing section for embedding the bit string obtained by the bit string generating means in a high frequency band of a quantization coefficient matrix of an adjacent block are provided. There is. According to this feature, generation and embedding of watermark information can be performed efficiently.

[0009]

DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS The present invention will be described below in detail with reference to the drawings. FIG. 1 is a block diagram showing a schematic configuration of an embodiment of the present invention.

In FIG. 1, an input image signal 1 is divided into a prediction signal subtracter 2, a motion estimation motion compensator 8, a switch 10,
And sent to the control unit 12. The motion-compensated prediction signal a is subtracted from the input image signal 1 by the prediction signal subtractor 2 and sent to a DCT unit (discrete cosine transform unit) 3 as a prediction error signal. The prediction error signal is orthogonally transformed by the DCT unit 3 and quantized by the quantization unit 4 in order to obtain high coding efficiency. Thereafter, the data is sent to the watermark embedding processing unit 15, undergoes watermark information embedding processing described later in detail, and is converted by the variable length coding unit 16 into a variable length code such as a Huffman code.
Then, after being temporarily stored in the buffer 17, it is output as a bit stream. The quantization unit 4 includes the buffer 14
Calculate the quantization step for the next block according to the rate control signal from.

Further, in order to use the same prediction signal as that on the decoding side, the quantized coefficients obtained by the quantization
, And the inverse DCT unit 6 locally decodes the prediction error signal. This prediction error signal is supplied to the local decoding adder 7
, And is added to the motion-compensated prediction signal a restored by the motion-estimated motion compensator 8, and sent to the motion-estimated motion compensator 8.

On the other hand, a random number generator 11 generates a random number for determining a block from which watermark information is to be extracted. Specifically, a block (8 × 8 pixels) in one frame of an image
2, numbers 0 to NB -1 (where NB is the number of blocks per frame) as shown in FIG.
The random number generator 11 generates 64 integers in the interval from 0 to NB -1. The control unit 12 controls the 6 obtained by the random number generator 11
The processing shown in the flowchart of FIG. 3 is performed based on the four integers.

That is, a watermark information embedding block determination process in step S1, a watermark information generation process in step S2, and a watermark information embedding process in step S3 are performed. In step S4, it is determined whether or not the processing of one frame of the input image 1 has been completed. When the determination is negative, 1 is added to i in step S5, and in step S2,
The process of S3 is repeated. If affirmative, the process ends.

Next, the details of the processing in step S1 will be described.
This will be described with reference to the flowchart of FIG. Step S
In step S11, the number of blocks is set to i = 0, and step S12
Then, it is determined whether or not the random number Bi has been input from the random number generator 11. If this determination is affirmative, step S
Proceeding to 13, it is determined whether the obtained random number Bi is an even number. If this judgment is affirmative (Bi is even),
Proceeding to step S14, the watermark information embedded block Ci is determined to be a (Bi + 1) block. On the other hand, if not (Bi is an odd number), the process proceeds to step S15,
The watermark information embedded block Ci is determined to be a (Bi -1) block. In step S16, it is determined whether or not i <NB-1. If the determination is affirmative, the process proceeds to step S17, where 1 is added to i. Then, returning to step S12, it is determined whether or not the next random number B1 has been obtained. The above processing is repeatedly performed, and if the determination in step S16 is negative, the process proceeds to step S2.

FIG. 5 is an explanatory diagram of the processing in steps S13 to S15. As shown in FIG. 7A, when the random number Bi is an even number, the watermark information embedding block C
Let i be an adjacent (Bi +1) block in the same macroblock (16 × 16 pixels), and as shown in FIG. 2B, if the random number Bi is an odd number, Let it be an adjacent (Bi -1) block in the macroblock. As will become clear from the explanation below,
If the embedding location of the watermark information is the same as that of the block Bi, the block Bi on the decoding side contains noise due to the watermark information, which may deteriorate the estimation of the SNR. When the processing of FIG. 4 is completed, 64 random numbers B0, B1,.
B63 and 64 watermark information embedding blocks C0, C
It is clear that 1, ..., C63 are determined.

Returning to FIG. 1, the description will be continued.
Monitors whether the input image signal 1 matches the block of the random numbers B0, B1,..., B63, and when they match, closes the switch unit 10. Therefore, the DCT unit 13 stores the random number B
Blocks corresponding to 0, B1,..., B63 are sent.

At this time, the control unit 12 determines in step S2
Is performed. The details will be described with reference to FIGS. In step S21, it is determined whether or not the input image block matches Bi. If the determination is affirmative, the switch unit 10 is closed and the block Bi is sent to the DCT unit 13 to perform the processing in step S22. That is, the i-th DCT component FBi [i] of the block Bi is calculated. Next, the process proceeds to step S23, where the i-th DCT component FBi [i] is converted into a bit string (dk) by the bit string converting unit 14. In the case of MPEG2, the DCT coefficient is represented using a maximum of 11 bits, but it is unlikely that these higher-order bits are changed due to an encoding error. Therefore, in the present invention, a sign bit is added to the lower 8 bits of the DCT coefficient, and a bit string of a total of 9 bits is used as watermark information. This bit string is represented as dk (k = 0, 1, 2,..., 8). Of these, d0 is a sign bit.

When the above processing is completed, step S
The watermark information embedding process of No. 3 is performed. Details of this processing will be described with reference to FIGS. Step S31
Then, the watermark embedding unit 15 selects the watermark information embedding block Ci. In step S32, the zigzag scan number of the quantization coefficient shown in FIG. 9 or the alternate scan number k = 0. In step S33, the quantization level in the high frequency range is manipulated, and processing for embedding watermark information in this region is performed. Specifically, d
It is determined whether the k = 0 and the (k + 55) th quantization coefficient = 0 or the dk = 1 and the (k + 55) th quantization coefficient ≠ 0 holds. If this determination is negative, the process proceeds to step S34, where the (k + 55) th quantized coefficient is set to dk. In step S35, it is determined whether or not k <9 is satisfied. When the determination is affirmative, the process proceeds to step S36, and 1 is added to k. And
The process of step S33 is performed again. The above processing is repeatedly performed, and if the determination in step S35 is negative, the processing returns.

By this processing, in the case of zigzag scanning, the 55th to 63rd positions of the quantization coefficient matrix of FIG. 9A are used, and in the case of the alternate scanning, the 55th to 63rd positions of the quantization coefficient matrix of FIG. Watermark information is embedded at the position of. The data to be embedded is 0 regardless of the value of QF [k + 55] when dk = 0, and QF [k + 55] when dk = 1 when the value of QF [k + 55] is 0. k + 55] = 1, and when the value of QF [k + 55] is a value other than 0, that value is used as it is.

The block in which the watermark information is embedded as described above is subjected to processing such as Huffman coding in the variable length coding unit 16, temporarily stored in the buffer 17, and then transmitted to the line at a predetermined transmission rate. Is done.

As described above, when the generation and embedding of the watermark information of one block is completed, the determination in step S4 of FIG. 3 is performed, and if this determination is negative, 1 is set to i.
Is added and the process returns to step S2. Then, the process proceeds to the process of determining the next watermark information embedded block. Blocks that do not correspond to the watermark information embedding block are simply passed through the watermark embedding processing unit 15 and sent to the variable length coding unit 16.

Next, the configuration of an embodiment of the decoding unit of the present invention will be described with reference to FIG. The bit stream 30 input from the encoding side is decoded by the variable length decoding unit 31 and input to the inverse quantization unit 32. The image signal that has been inversely quantized by the inverse quantization unit 32 is subjected to inverse DCT processing by an inverse DCT unit 33, and a motion compensation prediction signal from a motion compensation unit 35 is added by an adder 34 and output as a decoded image. . A general-purpose decoding device can be used.

In this embodiment, the blocks in which the watermark information is embedded are 64 blocks in one frame, and the blocks are embedded in the high frequency range of the quantization coefficient of the block.
Deterioration of the image quality of the decoded image is sufficiently small, and can be set as an allowable range of the deterioration. For this reason, a decoded image with good image quality can be obtained in parallel with the decoded image SNR estimation processing described below.

The random number generator 42 uses the same random number generation algorithm as the encoding side and shares random number seeds,
The random number can be specified without sending information on the watermark position from the encoding side. If the information to be shared is, for example, information of a specific few bits in the sequence header, the embedding position can be synchronized on the transmitting and receiving sides without obtaining special information from the outside.

When the control unit 43 obtains 64 random numbers Bi (i = 0, 1,..., 63) from the random number generator 42, the control unit 43 performs the same processing as that on the encoding side (see FIG. 4) to obtain one of the images. From the blocks in the frame, 64 watermark information embedded blocks C0, C1,..., C63 are determined. Control unit 43
Controls the switch unit 36 by the random number Bi and controls the switch unit 39 by the watermark information embedding block Ci. The switch unit 36 is closed at the timing when the Bi block is decoded. Then, a Bi block is extracted from the decoded image, sent to the DCT unit 37 and subjected to DCT transform, and its i-th DCT component FBi '[i] is calculated. The i-th DCT component FBi '[i] is sent to the control unit 43 and stored in a storage unit (not shown). In addition, the i-th DCT
The component FBi '[i] is bit-converted into 11 bits by the bit string converting unit 38 and sent to the watermark information detecting unit 40.

Next, the switch unit 39 is closed at the timing when the Ci block is received, and the Ci block is extracted and sent to the watermark information detecting unit 40. The watermark information detecting section 40 attaches the upper 2 bits of data obtained from the bit string forming section 38 to the 8-bit watermark information detected from the Ci block, and complements the upper 2 bits of data deleted at the time of embedding the watermark information. I do. This complement is accurate because the upper two bits cannot be altered by coding errors. The watermark information detection unit 40 also
From the value of the quantized coefficient QF [55], a plus / minus sign is detected, and a sign is attached to the 10-bit watermark information. The signed 11-bit watermark information is converted to a 10
To be evolved. The deciphered watermark information is sent to the control unit 43 and stored in a storage unit (not shown) or the like.

When the above processing is performed on an image of one frame, the storage means of the control unit 43 stores a virtual block Vrecv [i] composed of the i-th DCT component of the 64 Bi blocks of the decoded image. , A virtual block Vorig composed of watermark information of the original image detected from the 64 Ci blocks. The control unit 43 controls the virtual block V
Recv [i] and the virtual block Vorig [i] are sent to the inverse DCT unit 44, and inverse DCT is performed. Thereby, the image signals vrecv (n) and vorig (n) of the virtual block are obtained, and the SNR estimating unit 4
In step 5, the SNR estimation value per frame is obtained by the equation (1) or (2) in FIG.

In the present embodiment, since the number of DCT coefficients to be embedded as a watermark is very small as compared with the entire frame, the estimation accuracy for each frame is considered not to be very large. Therefore, the inter-frame average of these estimated SNRs is obtained, and the value given by equation (3) in FIG. 13 is used as the estimated SNR, thereby improving the estimation accuracy.

FIG. 11 is a flowchart showing a specific example of the SNR estimation processing of the present embodiment. In this processing, the processing from step S51 for reproducing embedded data from the processing block Ci and the processing from step S71 for reproducing data corresponding to the embedded data from the decoded image proceed in parallel.

First, in step S51, the number of blocks i
= 0, and it is determined in step S52 whether or not the processing block Ci has been received. If this determination is affirmative, the process proceeds to step S53 where watermark information WM = 0 is set. In step S54, k = 1 is set. In step S55, the 56th value QF [k + 55] of the quantization coefficient is set to 0.
Is determined. When this determination is negative, dk = 1 is determined in step S56, and when affirmative, dk = 0 is determined in step S57. Step S5
At 8, it is determined whether or not k <9 holds. If not, the process proceeds to step S59 to add 1 to k. The above processing is repeated, and if the determination in step S58 becomes positive, it means that the 8-bit watermark information has been reproduced from the processing block Ci.

In step S60, a later-described step S60
The upper two bits d'9 and d'10 of the watermark information are obtained from 74 and added to the upper part of the 8-bit watermark information. Then, in step S61, the watermark information dk of 10 bits in total is deciphered (= WM) into 10 bits. In step S62, the fifty-fifth quantization coefficient representing the sign
It is determined whether or not the 0th value is 0. If it is not 0, the process proceeds to step S63, and the 10-decoded watermark information WM
With a negative sign. On the other hand, if it is 0 (positive sign),
Proceed to step S64. In step S64, the ith value Vorig [i] of the virtual block is set to WM. In step S65, it is determined whether all the watermark information has been obtained. If the determination is negative, the process proceeds to step S66, where 1 is added to i. Subsequently, the process returns to step S52 and the above-described operation is repeated.

On the other hand, in the processing of the decoded image, i = 0 is set in step S71, and it is determined in step S72 whether or not the processing block of the decoded image has become Bi. If this determination is affirmative, step S73 is performed. And the i-th DCT component Fbi '[i] of the block Bi is calculated. In step S74, the i-th DCT component FBi '[i] is converted into a bit string. The value d′ k converted into a bit string is 1 including the sign.
1-bit data, but the upper 2 bits (d'9, d'1
0) is sent to step S60 as described above. In step S75, the i-th DCT component FBi '[i] is Vrecv
[i], and in step S76, it is determined whether or not all the watermark information has been obtained. If this determination is negative, the process proceeds to step S77, where 1 is added to i, and returns to step S72 to repeat the above-described processing.

As a result of the above operation, when the determinations in steps S65 and S78 are affirmative, 64 pieces of watermark information information have been obtained as DCT coefficients. Proceeding to steps S67 and S78, these watermark information
Each of them is subjected to inverse DCT processing, and is converted into signals vorig [n] and vrecv [n], which are the levels of the image signal.
Is subjected to SNR estimation processing. This process is the same as that of FIG.
This is performed using the equations (1) to (3).

As is clear from the above description, according to the present embodiment, a part of information relating to the pixel value of the original image is given as watermark information, transmitted to the receiving side and faithfully reproduced.
Since the SNR is estimated by comparing with the corresponding data of the decoded image, the SNR of the decoded image can be accurately estimated.

FIG. 12 is a block diagram showing a second embodiment of the present invention. In this embodiment, the watermark embedding processing unit 15 is arranged in the encoding loop of the encoding device, and the watermark information embedded by the watermark embedding processing unit 15 participates in the motion estimation as compared with that of FIG. There is a characteristic in the point that it did. According to the present embodiment, since motion estimation is performed in consideration of the watermark information, matching between the encoding device and the decoding device can be achieved, and the quality of a decoded image can be improved. Further, a general-purpose decoding device can be used.

In the following description, 64 blocks in one frame are used as the target blocks for extracting the watermark information. However, the present invention is not limited to this. Can be. As for the estimated SNR at this time, it is preferable to apply the above equation (2) to the average of the estimated MSE of the above equation (1).

[0037]

As is apparent from the above description, according to the present invention, the transmitting side embeds the DCT coefficient of the block Bi in the adjacent block Ci as watermark information and the receiving side embeds the DCT coefficient of the block Bi and the adjacent block Ci. And the SNR of the decoded image is estimated based on the result, so that the block Bi does not include noise due to the watermark information, and the estimation accuracy of the absolute image quality evaluation increases. Further, since the watermark information is converted into a bit string and embedded in the high frequency band of the quantization coefficient matrix, the deterioration of the decoded image due to the embedding of the watermark information is small.

Further, since the watermark embedding unit is arranged in the encoding loop of the encoding device, the encoding device and the decoding device can be matched, and the quality of the decoded image can be improved. A general-purpose decoding device can be used.

[Brief description of the drawings]

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

FIG. 2 is an explanatory diagram of numbers assigned to blocks in one frame of an image.

FIG. 3 is a flowchart illustrating an outline of watermark information embedding processing.

FIG. 4 is a flowchart showing details of step S1 in FIG. 3;

FIG. 5 is an explanatory diagram of a watermark information embedding block.

FIG. 6 is a flowchart showing details of step S2 in FIG. 3;

FIG. 7 is an explanatory diagram of the processing in FIG. 6;

FIG. 8 is a flowchart showing details of step S3 in FIG. 3;

FIG. 9 is an explanatory diagram of the processing in FIG. 9;

FIG. 10 is a block diagram illustrating a configuration of a decoding device according to an embodiment of the present invention.

FIG. 11 is a flowchart showing the operation of the decoding device.

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

FIG. 13 is a diagram showing a mathematical expression.

[Explanation of symbols]

1: input image signal, 3, 13, 37: DCT unit, 4: quantization unit, 5, 32: inverse quantization unit, 6, 33, 44 ... inverse D
CT unit, 8: motion estimation / motion compensation unit, 11, 42: random number generator, 12, 43: control unit, 14, 38: bit string conversion unit, 15: watermark embedding processing unit, 16, 31: variable length coding Unit, 40: watermark information detecting unit, 45: SNR estimating unit.

Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat II (Reference) H04N 1/41 H04N 7/08 Z 5J064 7/30 7/133 Z 5J104 17/00 9A001 (72) Inventor Shuichi Matsumoto Saitama 2-1-15 Ohara, Kamifukuoka-shi, Japan F-term in K.D. Research Institute Co., Ltd. (reference) 5C076 AA02 AA14 BA06 5C078 AA04 BA23 BA57 CA00 DA00 DA01 DA02 DB17 DB18 5J064 AA00 BA09 BA15 BA16 BC08 BC16 BC25 BD01 5J104 AA14 PA14 9A001 EE04 EE05 HH27 KK60 LL03

Claims (14)

[Claims] 1. Extracting a plurality of blocks per frame from an input image, extracting DCT coefficients from the plurality of blocks,
Encoding the extracted DCT coefficients into a bit string, embedding the obtained bit string in a quantization matrix of an adjacent block of the plurality of blocks, and transmitting the resultant signal; Method. 2. The digital watermark according to claim 1, wherein the plurality of blocks are randomly extracted, and the number of blocks to be extracted is a multiple of the number corresponding to the DCT coefficient of one block in total. A coding method for absolute image quality evaluation using. 3. The encoding for absolute image quality evaluation using a digital watermark according to claim 1, wherein the bit string is embedded in a high frequency band of a quantization matrix of an adjacent block of the plurality of blocks. Method. 4. The block according to claim 1, wherein when the block from which the DCT coefficient is extracted is an even number block or an odd number block, the block in which the bit string is embedded is an odd number block or an even number block, respectively. A coding method for absolute image quality evaluation using a digital watermark described in Crab. 5. The method according to claim 1, wherein one DCT coefficient is extracted per block, and components of the extracted DCT coefficient do not overlap in the plurality of blocks.
Encoding method for evaluating the absolute image quality using the digital watermark described in 1 above. 6. The method according to claim 6, wherein upper bits of the bit string are reduced,
4. The encoding method for evaluating absolute image quality using a digital watermark according to claim 3, wherein the encoding is performed in a high frequency range of a quantization matrix of adjacent blocks of the plurality of blocks. 7. A bit value of the bit string is 0 and a quantization coefficient is 0, or the bit value is 1 and a quantization coefficient is 0.
7. The absolute image quality evaluation using a digital watermark according to claim 6, wherein the quantization coefficient is left as it is when the condition that the integer is not satisfied, and the bit value is embedded when the condition is not satisfied. Encoding method. 8. A coded image signal including a block from which DCT coefficients are extracted according to claim 1 and 2 and an adjacent block in which a bit string is embedded, and a DCT coefficient of the block reproduced and reproduced, and the adjacent block. An absolute image quality evaluation method using a digital watermark, comprising estimating an SNR of a decoded image based on a DCT coefficient reproduced based on a bit string detected from the above and complemented with upper bits. 9. Each D of said block and adjacent blocks
9. The absolute image quality using digital watermark according to claim 8, wherein the CT coefficients are collected in multiples of the number corresponding to the DCT coefficients of one block, and are inversely DCTed to estimate the SNR of the decoded image. Evaluation method. 10. The digital watermark according to claim 8, wherein the DCT coefficient of the block is obtained by DCT of a decoded image, and the DCT coefficient of the adjacent block is obtained by deciphering the bit sequence. Absolute image quality evaluation method. 11. The method according to claim 8, wherein the upper bits are complemented by upper bits of a bit string obtained by converting a DCT coefficient obtained by DCT of the decoded image into a bit string. Absolute image quality evaluation method using digital watermark. 12. An absolute image quality evaluation method using a digital watermark, wherein an estimated SNR of one frame obtained in claim 9 is obtained for a plurality of frames, and an average value thereof is used as an estimated SNR. 13. A block extracting means for extracting a plurality of blocks from an input image signal, an orthogonal transform means for orthogonally transforming the blocks extracted by the block extracting means, and an orthogonal transform coefficient obtained by the orthogonal transform means. Using a digital watermark, comprising: a bit string forming means for forming a bit string; and a watermark information embedding processing section for embedding a bit string obtained by the bit string forming means in a high frequency range of a quantization coefficient matrix of an adjacent block. Encoding device for absolute image quality evaluation. 14. The encoding apparatus for evaluating absolute image quality using a digital watermark according to claim 13, wherein said watermark information embedding processing section is provided in an encoding loop.