JP2001060875A - Embedding device, digital camera and recording medium - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide an embedding device which causes reduced deterioration of picture quality and can embed voice data as much as possible. SOLUTION: An entropy decoding part 105 decodes with entropy a JPEG image and outputs a quantization DCT(discrete cosine transform) coefficient block. A selection part 106 selects 16 AC coefficients among the AC coefficients of low frequency which are included in the quantization DCT coefficient block and the AC coefficients having absolute value not smaller than its threshold, i.e., the AC coefficients having reduced influence of image deterioration. An embedding processing part 107 replaces the least significant bit of a prescribed number of AC coefficients with the partial compressed voice data of 16 bits which are inputted from a compressed voice input part 104.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an embedding apparatus for embedding audio data in an image and a digital camera having an embedding apparatus for embedding compressed audio data in a compressed image.

[0002]

2. Description of the Related Art Heretofore, there has been a demand for adding audio data to an image taken by a digital camera or the like. To meet this demand, there is a file format called FlashPix which handles image and audio data as one file. However, this FlashPix has a problem that the larger the data amount of audio data to be added, the larger the total data amount. If the amount of data is large, the storage and transmission efficiency will deteriorate.

Therefore, use of an embedding device can be considered as a means for preventing the total data amount from increasing. Since the embedding device performs embedding by replacing a part of the image with another information, the data amount of the image in which the different information is embedded is not different from the data amount before the embedding.

[0004]

However, the conventional embedding apparatus is intended to embed a small amount of data such as a signature or a mark as separate information, and to embed a large amount of data such as voice data. Is not taken into account. In view of the above points, the present invention has little degradation of image quality,
An object of the present invention is to provide an embedding device capable of embedding as much data as possible.

Another object of the present invention is to provide a digital camera capable of embedding audio data of about several seconds in a captured image with little deterioration of the image.

[0006]

In order to solve the above problems, an embedding apparatus according to the present invention provides a quantized DCT coefficient generated by performing discrete cosine transform (DCT) and quantization on an image. During the block, the AC coefficient and the absolute value of the low frequency are equal to or higher than the first threshold value.
Selecting means for selecting a predetermined number of AC coefficients from the C coefficients;
Replacement means for replacing the least significant bit of the predetermined number of AC coefficients selected by the selection means with audio data.

The selection means may include a first selection unit for selecting an AC coefficient having an absolute value equal to or greater than a first threshold value, and an AC coefficient selected by the first selection means when the number is less than a predetermined number. Is less than the first threshold and the lower frequency A
A second selection unit for selecting AC coefficients in order from the C coefficient until a predetermined number is reached. The selection unit further includes a third selection unit that selects an AC coefficient whose absolute value is equal to or greater than a second threshold value that is greater than the first threshold value, and the replacement unit is further selected by the third selection unit. The audio data is embedded in the second least significant bit of the AC coefficient.

According to the embedding method of the present invention, in a quantized DCT coefficient block generated by performing discrete cosine transform (DCT) and quantization on an image, a low-frequency AC coefficient and an absolute value are set to a first threshold value. From the above AC coefficient, a predetermined number of A
The method includes a selecting step of selecting a C coefficient, and a replacing step of replacing the least significant bit of the predetermined number of AC coefficients selected in the selecting step with audio data.

[0009] The selecting step includes a first selecting step of selecting an AC coefficient having an absolute value equal to or greater than a first threshold.
If the number of AC coefficients selected by the first selecting means is less than a predetermined number, the AC coefficients are selected until the predetermined number of AC coefficients are less than the first threshold and sequentially from lower frequency AC coefficients. 2 selecting steps. The digital camera of the present invention is a digital camera that embeds compressed audio data corresponding to audio data for several seconds into a compressed image, and obtains a quantized DCT coefficient block subjected to discrete cosine transform and quantization from the compressed image. Acquiring means; dividing means for dividing the compressed audio data into partial compressed audio data of a predetermined bit; wherein, in the obtained quantized DCT coefficient block, the low-frequency AC coefficient and the absolute value are equal to or more than a first threshold value. A
A selecting means for selecting the predetermined number of AC coefficients from the C coefficients; and a replacing means for replacing least significant bits of the predetermined number of AC coefficients selected by the selecting means with the partially compressed audio data.

A recording medium according to the present invention is a computer-readable recording medium in which a program for causing a computer to execute a process of embedding audio data in an image is recorded. In a quantized DCT coefficient block generated by performing transform (DCT) and quantization, a predetermined number of AC coefficients are selected from a low-frequency AC coefficient and an AC coefficient whose absolute value is equal to or greater than a first threshold. And a replacement step of replacing the least significant bit of the predetermined number of AC coefficients selected by the selection step with audio data.

The selecting step includes a first selecting step of selecting an AC coefficient having an absolute value equal to or greater than a first threshold value;
If the number of AC coefficients selected by the first selecting means is less than a predetermined number, the AC coefficients are selected until the predetermined number of AC coefficients are less than the first threshold and sequentially from lower frequency AC coefficients. It consists of two selection steps.

[0012]

(Embodiment 1) Hereinafter, a digital camera 1 which is an embodiment of an embedding device of the present invention will be described with reference to the drawings. In the present embodiment, the embedding device is provided inside the digital camera 1 as an embedding unit. (External Configuration of Digital Camera 1) FIGS. 1 and 2 are external views of the digital camera 1 on the front side and the rear side.

As shown in FIG. 1, the digital camera 1 includes a speaker 11 for reproducing audio data, a microphone 12 for recording audio data, and a lens 18 on the front, and an image display unit 13, an image and audio data reproducing instruction on the rear. Play button 21 for
a, 21b, recording button 22 for voice data recording instruction,
It has a viewfinder 17 and a shutter button 1 on the top.
4. A status display unit 15 for displaying a shutter speed, an aperture value and the like is provided, and a memory card insertion slot 19 for inserting a memory card 20 which is a kind of flash memory is provided on a side surface.

An operation example of the digital camera 1 will be briefly described below. When the user sets the imaging range using the viewfinder 17 or the image display unit 13 and presses the shutter button 14, the image captured through the lens 18 is internally encoded into a compressed image, and stored in the memory card 20. This memory card 20 can store several tens of compressed images.

When the user presses the record button 22, sound data for a predetermined time (about 10 seconds in the present embodiment) is collected from the microphone 12 from that point and is encoded to be compressed sound data. The compressed audio data is embedded in the compressed image selected by the user among the compressed images stored in the memory card 20. Hereinafter, a compressed image in which compressed audio data is embedded will be referred to as a compressed image with audio, and a non-embedded compressed image will be referred to as a compressed image without audio. It is called an image.

Each time the user presses the play buttons 21a and 21b, the compressed images stored in the memory card 20 are decoded one by one, and the images are displayed on the image display unit 13. If the compressed image with audio is decoded,
Simultaneously with the display of the decoded image, the compressed audio data is extracted, decoded, and reproduced from the speaker 11. (Schematic Configuration of Digital Camera 1) FIG. 3 is a schematic configuration diagram of the digital camera 1.

As shown in FIG. 1, the digital camera 1 includes an image encoding unit 3, an encoding memory 35, a memory card input / output unit 36, an audio encoding unit 4, an embedding unit 37, an audio decoding unit 5, and an image decoding unit. And an extraction unit 83. When the shutter button 14 is pressed, the image encoding unit 3 encodes an image captured through the lens 18 by the JPEG method to generate a compressed image without sound, and outputs the compressed image to the encoding memory 35.

The encoding memory 35 temporarily stores compressed images input and output among the image encoding unit 3, the memory card input / output unit 36, the embedding unit 37, and the image decoding unit 6.
The memory card input / output unit 36 writes the compressed image stored in the encoding memory 35 to the memory card 20 and reads the compressed image stored in the memory card 20 to the encoding memory 35.

When the recording button 22 is depressed, the audio encoding section 4 collects external audio data for about 10 seconds through the microphone 12, and outputs an IMA (Interactive Multimedia As
ADPCM (Adaptive Differential P)
CM), and the resulting compressed audio data is stored in a memory (audio memory
4).

The embedding section 37 embeds the compressed audio data in the compressed image without audio stored in the encoding memory 35 to generate a compressed image with audio, and returns the compressed image with audio to the encoding memory 35. The embedding section 37 is a main component of the present invention, and will be described later in detail. The image decoding unit 6 decodes the image from the compressed image stored in the encoding memory 35 by an operation reverse to the encoding performed by the image encoding unit 3, and displays the image on the image display unit 13.

The extraction unit 83 extracts compressed audio data from the compressed image with audio and outputs the extracted compressed audio data to the audio decoding unit 5 by performing an operation reverse to the embedding of the audio data by the embedding unit 37. The audio decoding unit 5 decodes the audio data from the compressed audio data output from the extraction unit 83 and reproduces the audio data from the speaker 11 by an operation reverse to the encoding performed by the audio encoding unit 4. (The image encoding unit 3, the audio encoding unit 4, the audio decoding unit 5,
FIG. 4 is a detailed configuration diagram of FIG. 3, and FIGS. 5 and 6 are detailed configuration diagrams partially showing FIG. The image encoding unit 3, the audio encoding unit 4, the audio decoding unit 5, and the image decoding unit 6 will be described below with reference to FIG. (Detailed Configuration of Image Encoding Unit 3) In FIG. 4, the image encoding unit 3 includes an imaging unit 31, a captured image memory 33, and an encoding unit 3.
4 (Imaging Unit 31) The imaging unit 31 includes a lens 18, a CCD (not shown), a color converter (not shown), and the like, and is obtained via the lens 18 and the CCD when the shutter button 14 is pressed. The RGB signal is converted into an image composed of YCrCb components by a color converter, and written into the captured image memory 33.

One image has a luminance component Y consisting of 1280 pixels × 960 lines, a total of 1,228,800 pixels, and 6
Color difference components Cr and Cb composed of 40 pixels × 480 lines (or 640 pixels × 960 lines thinned only in the horizontal direction), for a total of 307200 pixels (or 614400 pixels)
It is composed of (Captured Image Memory 33) The captured image memory 33 temporarily stores an image written by the imaging unit 31. (Encoding unit 34) The encoding unit 34 includes the captured image memory 33
Is compressed and encoded by the JPEG method for each block of 8 pixels × 8 lines, and the resulting compressed code string is written in the coding memory 35. A compressed code string for one screen corresponds to a compressed image without sound.

FIG. 7 shows the relationship between the luminance component Y for one screen and blocks. The luminance component Y is composed of 160 blocks in the horizontal direction and 120 blocks in the vertical direction, for a total of 19,200 blocks. Each block is composed of 8 pixels × 8 lines, that is, a total of 64 pixels. For example, in FIG.
Reference numeral 2 denotes one block included in the luminance component Y for one screen. Similarly, the color difference components Cr and Cb for one screen are also divided from 80 horizontal blocks and 60 vertical blocks, respectively, for a total of 4800 blocks (80 horizontal blocks, 120 vertical blocks, and a total of 9600 blocks for 640 pixels × 960 lines). Be composed.

FIG. 8A shows a specific example Yxy (x, y = 0 to 7; x, y represents a pixel position in a block) of a pixel of one block of the luminance component Y. Note that Yxy in the figure is obtained by subtracting 128 from the original signal value. This is a discrete cosine transform (DC).
This is for level-shifting the expected value of the DCT coefficient obtained by T) to 0. (Detailed Configuration of Encoding Unit 34) The encoding unit 34 includes a DCT unit 71, a quantization unit 72, and an entropy encoding unit 74, as shown in the configuration diagram of FIG. (DCT unit 71) The DCT unit 71
, The luminance component Y and the chrominance components Cr and Cb are sequentially read out for each block and DCT is performed to generate a DCT coefficient block Suv (u, v = 0 to 7) composed of DCT coefficients of 8 × 8 elements, and a quantization unit 72. Here, Suv refers to S00 representing a DC component as a DC coefficient, and DCT coefficients representing AC components other than S00 as an AC coefficient. Suv is u, v
The higher the value, the higher the frequency component.

FIG. 8B shows a DCT coefficient block Suv obtained by performing DCT on Yxy. In the figure, S00 = 823 is the DC coefficient, and the others are A
C coefficient. It can be seen that the value decreases as u and v increase, that is, as the frequency component increases. Note that the specific arithmetic expression of DCT is well-known, and thus description thereof is omitted. (Quantization Unit 72) The quantization unit 72 includes a quantization table Quv (u, v = 0 to 7) including quantization coefficients of 8 × 8 elements, and quantizes the DCT coefficient block Suv using this.
A quantized DCT coefficient block Ruv (u, v = 0 to 7) composed of 8 × 8 element quantized DCT coefficients is generated and output to the entropy coding unit 74.

The quantized DCT coefficient block Ruv is calculated as follows. (Equation 1) Ruv = round (Suv / Quv) Here, round () is a function meaning that the value in () is converted into an integer to the nearest integer.

In the JPEG system, the value of the quantization coefficient is not specified. Therefore, a value can be freely set for each application or each image. In general, the quantization coefficient is set to a larger value as the values of u and v become larger. The reason why the value of the quantization coefficient is increased as the frequency component increases is that if the high-frequency component that is hardly noticeable in deterioration is roughly quantized, the compression efficiency can be improved while protecting the image quality.

FIG. 8C shows a specific example of the quantization table Quv. FIG. 8D shows a quantized DCT coefficient block Ruv when the DCT coefficient block Suv shown in FIG. 8B is quantized by the quantization table Quv in FIG. 8C. According to this example, R10 = round (S10 / Q10) = round (-1)
35/4) = − 34. (Entropy coding unit 74) Entropy coding unit 7
4 entropy-encodes the quantized DCT coefficient block Ruv received from the quantization unit 71 to generate a compressed code string, and writes it to the code memory 35. A compressed code string for one screen corresponds to a compressed image. Since the entropy decoding is known, its description is omitted. (Detailed Configuration of Speech Encoding Unit 4) The speech encoding unit 4 includes a sound collection unit 41, an audio encoding unit 43, and an audio memory 4.
4 (Sound Collection Unit 41) The sound collection unit 41 includes the microphone 12, an amplifier (not shown), an AD conversion circuit (not shown), a quantization circuit (not shown), and the like, and the recording button 22 is pressed by the user. The external analog audio data for about 10 seconds from the time when the analog audio data is collected is converted into digital audio data by performing 11 kHz sampling, AD conversion, and the like.
Output to 3. (Audio encoding unit 43) Audio encoding unit 43
Converts the digital audio data into compressed audio data based on IMA ADPCM and outputs the compressed audio data to the audio memory 44. Since the IMA type ADPCM is publicly known, the description is omitted. (Audio Memory 44) The audio memory 44 stores the compressed audio data output by the audio encoding unit 43. (Detailed Configuration of Audio Decoding Unit 5) The audio decoding unit 5 includes an audio memory 54, an audio decoding unit 53, and an audio reproducing unit 51. (Audio Memory 54) The audio memory 54 temporarily stores the compressed audio data extracted from the compressed image by the extraction unit 83. (Audio Decoding Unit 53) Audio Decoding Unit 53
Decodes the digital audio data from the compressed audio data stored in the audio memory 54 based on the IMA type ADPCM, and outputs the digital audio data to the audio reproduction unit 51. (Audio Reproducing Unit 51) The audio reproducing unit 51 is composed of a DA conversion circuit (not shown), the speaker 11, and the like, and converts digital audio data decoded by the audio decoding unit 53 into analog audio data and reproduces the same. (Detailed Configuration of Image Decoding Unit 6) The image decoding unit 6 includes a decoding unit 62, a display image memory 61, and an image display unit 13. (Decoding Unit 62) The decoding unit 62 reads out the compressed image without sound or the compressed image with sound stored in the encoding memory 35, decodes the compressed image with the JPEG method, and stores the resulting image in the display image memory 61. Output.

FIG. 6 is a more detailed block diagram of the decoding section 62. In the figure, the decoding unit 62 includes an entropy decoding unit 84, an inverse quantization unit 82, and an inverse DCT unit 81. (Entropy decoding unit 84) Entropy decoding unit 8
4 generates a quantized DCT coefficient block Ruv (or R'uv) by entropy decoding the compressed image stored in the encoding memory 35, and outputs it to the extraction unit 83 and the inverse quantization unit 82. Here, the quantized DCT coefficient block that does not include the compressed audio data is identified as Ruv, and the quantized DCT coefficient block that includes the compressed audio data is identified as R'uv. (Inverse Quantization Unit 82) The inverse quantization unit 82 includes the same quantization table Quv as the quantization unit 71,
A DCT coefficient block S′uv is generated from the coefficient block Ruv (or R′uv) by inverse quantization shown in (Equation 2) and output to the inverse DCT unit 81. Here, the quantized DCT
The DCT coefficient block generated from the coefficient block Ruv or R'uv is identified as S'uv, and the DCT coefficient block before quantization is identified as Suv. (Equation 2) S′uv = Ruv (or R′uv) × Quv (Inverse DCT section 81) The inverse DCT section 81 performs inverse DCT to obtain a luminance component Y,
The color difference components Cr and Cb are decoded in block units and written into the display image memory 61. Since the inverse DCT is known, its description is omitted. (Display Image Memory 61) The display image memory 61 stores the luminance component Y and the chrominance component C decoded by the decoding unit 62.
The image composed of r and Cb is temporarily stored. (Image Display Unit 13) The image display unit 13 is configured by a liquid crystal display or the like, and displays an image stored in the display image memory 61. (Detailed Configuration of the Embedding Unit 37) The embedding unit 37 reads out the compressed image without sound stored in the encoding memory 35, performs entropy decoding, and performs quantization DCT coefficient block Ruv.
Return to Next, the embedding unit 37 performs quantization D of the luminance component Y.
For each CT coefficient block Ruv, the DC coefficient (R
(00), a predetermined number N of the quantized DCT coefficients (AC coefficients) other than the 63 quantized DCT coefficients are selected for embedding. In the present embodiment, the predetermined number N is set to 16. Further, the embedding unit 37 divides the compressed audio data into partial compressed audio data of N bits (ie, 16 bits). Finally, the embedding unit 37 embeds the partial compressed audio data one bit at a time in the least significant bit of the 16 quantized DCT coefficients for embedding so that each block corresponds to each partial compressed audio data. Here, embedding one bit of partially compressed audio data in the least significant bit of the quantized DCT coefficient means that the value of the least significant bit of the quantized DCT coefficient is replaced with one bit of partially compressed audio data.

Finally, the embedding unit 37 entropy-encodes the embedded quantized DCT coefficient block R'uv again and returns it to the encoding memory 35. FIG. 9 is a detailed configuration diagram of the embedding unit 37. In the figure, an embedding unit 37 includes a compressed image input unit 101, a determination value input unit 102,
It comprises an embedding amount input unit 103, a compressed speech input unit 104, an entropy decoding unit 105, a selection unit 106, an embedding processing unit 107, and an output unit 108.

The compressed image input section 101 reads out a compressed code string stored in the coding memory and outputs it to the entropy decoding section 105. The judgment value input unit 102 sets the judgment value J
Is stored in advance. The determination value J is determined by the quantization
Quantization DC for embedding from CT coefficient block Ruv
This is a threshold for determining the T coefficient. In the present embodiment, the judgment value J is 2. Quantization D of this judgment value 2 or more
CT coefficients are candidates for embedding.

The embedding amount input unit 103 has an embedding amount N
Is stored in advance. The embedding amount N is a value obtained by dividing the data amount of 38400 bytes of the compressed audio data corresponding to about 10 seconds by the total number of blocks of the luminance component and converting it into bits. In the present embodiment, N is 38400 bytes / 192.
00 block = 16 bits. Compressed voice input unit 10
Reference numeral 4 divides the compressed audio data stored in the audio memory 44 into the embedding amount N stored in the embedding amount input unit 103 and outputs the divided data to the embedding processing unit 107.

The entropy decoding unit 105 entropy decodes the compressed code string output from the compressed image input unit 101 and outputs a quantized DCT coefficient block Ruv of the luminance component to the selection unit 106. (Selection Unit 106) The selection unit 106 is a quantized DCT coefficient block Ruv output from the entropy decoding unit 105.
Each time, a total of N (16) quantized Ds out of 63 quantized DCT coefficients (AC coefficients) excluding the DC coefficient (R00)
The CT coefficient is selected for embedding, and the selection result is output to the embedding processing unit 107.

The selecting section 106 selects 8 × 8 embedding flags Euv (u, v) corresponding to the quantized DCT coefficient block Ruv.
= 0 to 7), and the selection result is recorded by setting an embedding flag Euv corresponding to the selected quantized DCT coefficient. More specifically, the embedding flag E
uv is initially set to off, and the selection unit 106 sets the selection flag corresponding to the quantized DCT coefficient selected for embedding to on. The selection flag corresponding to the quantized DCT coefficient not selected for embedding remains off.

The selecting section 106 selects a quantization DCT coefficient for embedding so that, when an image is decoded later, visual deterioration is minimized as compared with the original image before encoding. To do so, in the present embodiment, the selection unit 106
The absolute value of the quantized DCT coefficient is equal to or larger than the judgment value J,
An embedding quantized DCT coefficient is selected while giving priority to low frequency ones.

The reason for using the condition (1) is that a quantized DCT coefficient having a small absolute value has a larger error when a 1-bit value changes than a quantized DCT coefficient having a large absolute value. Therefore, embedding in a quantized DCT coefficient having a large absolute value causes less deterioration. The reason for using the condition is that the high-frequency quantized DCT coefficient is inversely quantized with a larger value than the low-frequency quantized DCT coefficient, as can be seen from the quantization table Quv in FIG. Is done. For this reason, the high-frequency quantized DCT coefficient is compared with a case where the low-frequency quantized DCT coefficient is embedded, and the DCT coefficient resulting from the inverse quantization is a DCT coefficient at the time of encoding.
The error increases as compared with the T coefficient. Therefore, the quantization DCT coefficient of a low frequency is less deteriorated when embedding is performed than the high frequency.

The reason for giving higher priority is that the higher the value, the less the deterioration when the value changes. FIG.
0 is a flowchart showing a selection process by the selection unit 106. In the same figure, the selection unit 106
One T coefficient (AC coefficient) is read, and it is determined whether or not its absolute value is equal to or greater than a determination value J (= 2) (steps 11 and 12). This determination may be made based on whether or not there is 1 in the second or more least significant bit value of the quantized DCT coefficient. That is, the selection unit 106 sets 1 to 1 from the lowest bit value of the absolute value of the quantized DCT coefficient.
If there is more than one, the absolute value of the quantized DCT coefficient is a value of 2 or more, and if there is no 1, the absolute value of the quantized DCT coefficient is smaller than 2.

If the result of the determination indicates that the absolute value is equal to or greater than the determination value J, the corresponding embedding flag is set and the variable C is set to 1
Is added (steps 13 and 14). The variable C indicates the number of set embedding flags, that is, the number of quantized DCT coefficients selected for embedding. As described above, the selecting unit 106 scans the quantized DCT coefficient block in the zigzag order, and executes the 63 quantized DCT coefficients (AC coefficients).
, The processing of steps 11 to 14 is repeated. This processing ends when the variable C is equal to or greater than the embedding amount N or when the processing of steps 11 to 14 is performed for all of the 63 quantized DCT coefficients (AC coefficients).
Proceed to 6.

In step 16, the selection unit 106 determines whether or not the variable C is smaller than the embedding amount N. If the result of the determination is that the variable C is smaller than the embedding amount N, it means that the number of quantized DCT coefficients selected for embedding has not reached the data amount of the partially compressed audio data. From the low-frequency side, in the zigzag scan order, the embedding flags are set by selecting (embedding amount N-variable C) pieces for embedding from among those whose embedding flags are turned off (step 17).

The selecting unit 106 selects N embedding quantized DCT coefficients by the above procedure. (Selection Processing Example 1) FIG.
The quantization D selected for embedding when the selection processing is performed on the quantized DCT coefficient block Ruv in (d).
The CT coefficients are shown by circles.

The selector 106 determines whether each quantized DCT coefficient (AC coefficient) is equal to or greater than the judgment value J while scanning the quantized DCT coefficient block Ruv in a zigzag order. In the example shown in FIG.
, -19, 30,... Are sequentially selected for embedding, and the quantization D
Since the number of CT coefficients has reached the embedding amount N, the selection process has been completed.

FIG. 4B shows that the selecting unit 106 is the same as FIG.
Indicates an embedding flag Euv of the selection result when the selection process is performed for the. In FIG. 9B, 1 indicates that the flag is set, and the quantization D corresponding to the position is set.
Indicates that the CT coefficient has been selected for embedding. Further, 0 indicates that the flag is not set, and indicates that the quantized DCT coefficient corresponding to that position is not selected for embedding. (Selection Processing Example 2) FIG.
FIG. 1 shows a quantized DCT coefficient block different from (d).
As a result of repeating steps 11 to 15 of 0, the quantized DCT coefficients selected for embedding are indicated by circles. As shown in the drawing, the selection unit 106 selects the embedding order in the order of 10, -11, -12, 5, 5, 12,... In the zigzag scanning order, and scans to the end of the quantized DCT coefficient block. ing.

FIG. 12B shows an embedding flag corresponding to FIG. As shown in FIG. 7B, the number of embedding flags set is thirteen, which is smaller than the embedding amount N (= 16). Then, the selection unit 106 performs the process of step 17. More specifically, the embedding flag shown in FIG.
= Set three unset embedding flags and set a total of 16 embedding flags.
The embedding flag surrounded by a circle in FIG.
It is newly set by the processing in step 17.

FIG. 12C shows an embedding flag of a final selection result by the selection processing of the selection unit 106. (Embedding Processing Unit 107) The embedding processing unit 107 embeds partially compressed audio data into the 16 quantized DCT coefficients selected for embedding by the selecting unit 106 for each quantized DCT coefficient block Ruv.

More specifically, the embedding processing unit 107 scans the embedding flags output from the selecting unit 106 in a zigzag scan order, and searches for the set embedding flags. When finding the embedding flag that has been set, the embedding processing unit 107 converts the least significant bit of the quantized DCT coefficient corresponding to the embedding flag into the compressed audio input unit 1.
04 is changed to 1 bit of the partially compressed audio data. The embedding processing unit 107 repeats this operation for the set 16 flags. In this way, the embedding processing unit 107 embeds one bit at a time from the beginning of the partially compressed audio data in the least significant bit of the quantized DCT coefficient for embedding, and outputs the embedded quantized DCT coefficient block to the output unit 108. I do. (Example of Embedding Process) FIG. 13A shows an example of the partially compressed audio data. The embedding processor 107 embeds one bit at a time from the beginning of the partially compressed audio data into the quantized DCT coefficient selected for embedding.

FIG. 11B shows the result when the embedding processing unit 107 embeds the partially compressed audio data of FIG. 10A into the quantized DCT coefficient block of FIG. The quantized DCT coefficients circled in FIG. 13B indicate that the partially compressed audio data is embedded. The embedding processing unit 107 scans the embedding flag of FIG. 12C in a zigzag scan order to search for the set embedding flag, and when found, embeds one bit of the partially compressed audio data in the quantized DCT coefficient corresponding to the flag. . (Output unit 108) The output unit 108
7 performs the same entropy coding as the entropy coding unit 74 on the quantized DCT coefficient block embedded therein, and stores the resulting compressed code string in the coding memory 3.
5 is output. This compressed code string is a compressed image with sound.

As described above, the embedding 37 embeds the partially compressed audio data in the quantized DCT coefficient block obtained by decoding the compressed code string, encodes the compressed code string again, and outputs it to the coding memory. By repeating the process, compressed audio data of about 10 seconds is embedded in one compressed image without audio. Thus, the embedding section 37
Performs embedding of partially compressed audio data in the least significant bit of 16 quantized DCT coefficients that do not affect image degradation for each block, so that 2 bytes × 192
The embedding of 00 blocks = 38400 bytes is performed, so that more compressed audio data can be embedded with less image deterioration. (Detailed Configuration of Extraction Unit 83) FIG. 14 is a detailed configuration diagram of the extraction unit 83.

In the figure, the extraction unit 83 is composed of an identification unit 83
1. It is composed of an extraction processing unit 832. Identification unit 831
When the quantized DCT coefficient block Ruv or R'uv is output from the entropy decoding unit 84, an embedding flag is generated by the same processing as in the flowchart of FIG. Here, the identification unit 831 generates an embedding flag regardless of whether the quantized DCT coefficient block is Ruv or R'uv. That is, the embedding flag is generated regardless of whether the compressed image has no sound or has sound. As a result, if it is the quantized DCT coefficient block R'uv of the compressed image with sound, the identification unit 831 restores the same embedded flag created by the selecting unit 106, and the quantized DCT coefficient block of the compressed image without sound. If it is Ruv, the identification unit 831 generates an embedded flag in which all flags are off.

The extraction processing unit 832 refers to the embedding flags restored by the identification unit 831 in a zigzag scan order, extracts the least significant bit of the quantized DCT coefficient corresponding to the embedding flag set to ON, and outputs Output to the memory 54. As described above, the extraction unit 83
Is a quantization DC in which the partially compressed audio data is embedded by performing a process reverse to the selection process and the embedding process.
The compressed audio data of about 10 seconds is extracted by extracting the partially compressed audio data from the T coefficient block and performing this for all the quantized DCT coefficient blocks. Embodiment 2 Hereinafter, a digital camera 2 according to Embodiment 2 of the present invention will be described.

The digital camera 2 embeds the compressed audio data in the least significant bit of the quantized DCT coefficient in the same manner as the digital camera 1 and also adds the second and third bits from the least significant bit of the quantized DCT coefficient. By embedding compressed audio data, more compressed audio data can be embedded in one image than in the first embodiment. Its configuration is different from that of the digital camera 1 shown in FIG.
An embedding unit 47 instead of the embedding unit 37 and the extraction unit 83
And an extraction unit 93.

Hereinafter, the embedding section 47 and the extracting section 93 will be described. (Embedding Unit 47) FIG. 15 is a detailed configuration diagram of the embedding unit 47. In the figure, an embedding unit 47 includes a compressed image input unit 101, a determination value input unit 202, and an embedding amount input unit 1.
03, compressed speech input unit 204, entropy decoding unit 1
05, selection unit 206, embedding processing unit 207, output unit 1
08.

In the figure, the components having the same reference numerals as those of the embedding unit 37 in FIG. 9 have the same functions, and thus the description thereof will be omitted, and the components having the different reference numerals will be described below. The determination value input unit 202 stores the determination values J, J2, and J3 in advance. In the present embodiment, J, J2, and J3 are 2, 4, and 8, respectively.
It is. The compressed voice input unit 204 is provided in the audio memory 4
4 is output to the embedding processing unit 207 by dividing the compressed audio data into units of embedding amount. The embedding amount is
It is different for each quantized DCT coefficient block.
06.

The selection unit 206 is the same as the first embodiment shown in FIG.
A quantization DCT coefficient to be embedded in the least significant bit is selected by a selection process indicated by “0”. This processing is performed in the first embodiment.
Therefore, the description is omitted. In addition to the above selection processing, the selection unit 206 selects a quantized DCT coefficient to be embedded in the second and third bits.

The selecting section 206 classifies each quantized DCT coefficient according to the magnitude of its absolute value, and embeds the number of quantized DCT coefficients to be embedded according to the number of quantized DCT coefficients belonging to each class and the embedding. Is determined. FIG. 16 is a flowchart illustrating the logic for determining the number of quantized DCT coefficients to be embedded and the bit position to be embedded by the selection unit 206.

In the figure, the selection unit 206 first determines the number C2 of quantized DCT coefficients whose absolute value is 4 or more and the quantized DCT coefficients whose absolute value is 8 or more in 63 AC coefficients in one quantized DCT coefficient block. The number C3 of DCT coefficients is counted (step 21). Next, the selection unit 206 makes the following determination according to the values of C2 and C3. When C3 is 8 or more, that is, when the number of quantized DCT coefficients whose absolute value is 8 or more is 8 or more in the quantized DCT coefficient block (step 2)
2), the selection unit 206 performs four quantizations D with absolute values of 8 or more.
The third bit from the lower order of the CT coefficient is determined to be embedded, and the second bit from the lower order is determined to be embedded for eight quantized DCT coefficients whose absolute values are 4 or more (step 23).

When C3 is less than 8 and C2 is 8 or more, that is, the number of quantized DCT coefficients having an absolute value of 8 or more in a quantized DCT coefficient block is less than 8 and the absolute value is 4 or more. When the number of quantized DCT coefficients is eight or more (step 24), the selection unit 206 determines that the absolute value is
The second least significant bit of the quantized DCT coefficients is determined to be embedded (step 25).

When C2 is 4 or more and less than 8, that is, the quantized DCT whose absolute value is 4 or more in the quantized DCT coefficient block
When the number of coefficients is 4 or more and less than 8 (step 2
6), the selector 206 selects four quantized Ds whose absolute values are 4 or more.
The second bit from the bottom of the CT coefficient is determined to be embedded (step 27). The selecting unit 206 has an embedding flag Euv similar to that of the first embodiment, and also includes an embedding flag E2uv (u, v = 0 to 2) for the second and third bits from the lower order.
7) and an embedding flag E3uv (u, v = 0 to 7), and the embedding flags E2uv and E3 according to the result of the above determination.
Set uv.

More specifically, the result of the determination is
, The selection unit 206 performs quantization DC
Scan the T coefficient (AC coefficient) in zigzag scan order,
Four quantized DCT coefficients having an absolute value of 8 or more are selected, and the corresponding embedding flag E3uv is set. Similarly, the quantized DCT coefficients (AC coefficients) excluding the DC coefficients are scanned in a zigzag scan order. Eight quantized DCT coefficients having an absolute value of 4 or more are selected, and the corresponding embedding flag E2uv is set.

If the result of the determination is step 25,
The selection unit 206 outputs the quantized DCT coefficients (AC
Are scanned in the zigzag scan order, eight quantized DCT coefficients having an absolute value of 4 or more are selected, and the corresponding embedding flag E2uv is set. Embedded flag E3
uv is not set. If the determination result is step 27, the selection unit 206 scans the quantized DCT coefficients (AC coefficients) excluding the DC coefficients in a zigzag scan order, selects four quantized DCT coefficients having an absolute value of 4 or more, and The corresponding embedding flag E2uv is set. The embedding flag E3uv is not set.

For example, when the decision processing is performed on the quantized DCT coefficient block shown in FIG.
The above AC coefficients are 10, -11, -12, 5, 5, 1
Since 2, -7, -7 and 4 are 9 pieces, C2 = 9 and the absolute value is 8
Since the above-mentioned AC coefficients are 10, -11, -12, and 12, C3 = 4 (step 21). Therefore, since the determination result corresponds to step 25, the second bit from the lower order of the eight quantized DCT coefficients whose absolute values are 4 or more from the quantized DCT coefficient block in FIG. Is done. FIG. 17 shows the embedding flag E2uv set according to the determination result. As shown in the figure, eight quantized DCT coefficients whose absolute values are 4 or more are selected in zigzag scan order, and the corresponding embedding flag E2uv
Is set. According to the determination result of step 25, no embedding flag E3uv is set.

Further, the selection unit 206 notifies the compressed voice input unit 204 of the embedding amount including the least significant bit and the second and third bits from the lowest. Specifically, when the decision result is step 23, the embedding in the second bit from the lower order is 8 bits, the embedding in the third bit from the lower order is 4 bits, and the embedding in the least significant bit is 16 bits. , The selector 206 notifies the total 28 of 12 and 16 as the embedding amount.

When the decision result is step 25, the embedding in the second bit from the lower order is 8 bits, and the embedding in the least significant bit is 16 bits. Notify as. If the determination result is step 27, the embedding in the second bit from the lower order is 4 bits, and the embedding in the least significant bit is 1
Since the number of bits is 6 bits, the selection unit 206 outputs a total of 2
0 is notified as the embedding amount.

The embedding processing unit 207 embeds the partially compressed audio data in the least significant bit, the second least significant bit, and the least significant third bit of the quantized DCT coefficient based on the embedding flags Euv, E2uv, and E3uv. Specifically, the embedding processing unit 207 performs embedding in the least significant bit based on the embedding flag Euv in the same manner as in the first embodiment. Since this is the same as in the first embodiment, the description is omitted.

Next, the embedding processing section 207 scans the embedding flag E2uv in a zigzag scan order, searches for the set embedding flag, and if found, fetches the second lower bit of the corresponding quantized DCT coefficient into the compressed audio input signal. It is changed to one bit of the partially compressed audio data input from the unit 204. When the scanning of the embedding flag E2uv is completed, the embedding processing unit 207 similarly scans the embedding flag E3uv in the zigzag scanning order, and performs quantization DCT corresponding to the set embedding flag.
The third bit from the lower order of the coefficient is changed to 1 bit of the partially compressed audio data.

As described above, the embedding section 47 selects the quantized DCT coefficient and its bit position which do not affect the image degradation for each block and embeds the partially compressed audio data. ４８48 Kbytes of compressed audio data will be embedded. Extraction unit 93
When the quantized DCT coefficient block is output from the entropy decoding unit 84, the embedding flags Euv, E2uv, and E3uv are restored by the same method as the processing performed by the selecting unit 206, and the quantized DCT coefficient is Partially compressed audio data is extracted from the least significant bit, the second least significant bit, and the third least significant bit of the coefficient, and is output to the audio memory 54.

As described above, the extraction unit 93 is provided with the embedding unit 47.
By performing the reverse operation, the partially compressed audio data can be extracted. As described above, Embodiments 1 and 2 of the present invention
However, the present invention is not limited to the first and second embodiments, and may be as follows. (1) Although the embedding unit 37 is provided inside the digital camera in the first embodiment, the embedding unit 37 may be configured as a single unit without being provided inside the digital camera. The same applies to the embedding part 47 of the second embodiment.

The embedding sections 37 and 47 may be configured inside a device capable of image processing such as a personal computer. (2) In the first embodiment, the embedding unit 37 quantizes the compressed image once encoded by entropy decoding.
Although the embedding is performed by the procedure of embedding the partially compressed audio data after returning to the CT coefficient block, the embedding may be performed at the stage of encoding by the encoding unit 34. More specifically, the embedding unit 37 uses the DCT unit 71 and the quantization unit 72 in the encoding unit 34
The embedding is performed on the quantized DCT coefficient block after being subjected to T and quantization and before being encoded by the entropy encoding unit 74. In this case, the embedding unit 3
7 has an entropy decoding unit 105 and an output unit 108
Becomes unnecessary. (3) The data amount of the compressed image, the determination value, the embedding amount, the quantization table Quv, and the like are not limited to the values described in the first and second embodiments.

For example, the judgment value J may be 3 or 4. However, it is desirable to use 2 ⁿ (n is a natural number). The reason is that the selection unit 106 selects (n
This is because whether or not the coefficient is ²ⁿ or more can be easily determined by determining whether or not there is 1 in the (+1) th or more bit value. Further, a configuration may be adopted in which a plurality of different quantization tables are provided, and the determination value and the embedding amount may be changed according to the quantization tables.

The embedding amount in the first, second, and third bits from the lower order of the quantized DCT coefficient may be made larger than that in the second embodiment in consideration of the balance between the image deterioration and the embedding amount. For example, if the value of the quantization table is reduced as a whole to lower the compression ratio, the amount of embedding may be increased such as 30 bits for the first bit, 16 bits for the second bit, and 8 bits for the third bit. It is possible. (4) entropy encoding section 74 and output section 108, entropy decoding section 105 and entropy decoding section 84,
In the digital camera, two components having the same function, such as the photographed image memory 33 and the display image memory 61, the audio memory 44 and the audio memory 54, may be omitted and one may be used alone. (5) In the first embodiment, the digital camera 1 is configured to uniformly embed partially compressed audio data in each quantized DCT coefficient. According to this configuration, for a block in which the number of quantized DCT coefficients equal to or greater than the determination value J is smaller than the embedding amount N, embedding is performed also for quantized DCT coefficients having a value smaller than the determination value J. There is a problem that it will be deteriorated more than other blocks. Such a problem may be dealt with by the following method.
That is, the selecting unit 106 first sets, for each quantized DCT coefficient,
Among the 63 AC coefficients excluding the DC coefficient, the number of coefficients whose absolute value is equal to or greater than the determination value J is counted. Next, the selection unit 1
In step 06, it is determined whether the number of counted coefficients is equal to or larger than the value of the embedding amount N. As a result of the determination, when the embedding amount is equal to or more than the value of the embedding amount N, it is determined that embedding is performed in the quantized DCT coefficient block, and the quantized DCT coefficient blocks having the judgment value J or more are selected for embedding in zigzag scan order. To set the embedded flag. As a result of the judgment,
When the value is smaller than the value of the embedding amount N, the quantization D
It is determined that embedding is not performed in the CT coefficient block, and the embedding flag is not set. Embedding processing unit 10
7 embeds the partially compressed audio data only in the block having the embedding flag set by the selection unit 106. Thereby, the data amount of the compressed audio data that can be embedded is smaller than that of the configuration of the first embodiment, but there is an effect that the image quality can be protected more than the configuration of the first embodiment. (6) The function of each component of the embedding unit 37 may be programmed and recorded in the ROM, and may be realized by a microcomputer including a CPU, a RAM, and a ROM. More specifically, the CPU reads out the program from the ROM and executes the program, so that the low-frequency AC coefficient and the AC value whose absolute value is equal to or more than the determination value 2 are selected from the quantized DCT coefficient block Ruv.
A selection step of selecting 16 AC coefficients from the coefficients and a replacement step of replacing the least significant bit of the 16 AC coefficients selected by the selection process with 16-bit partially compressed audio data are performed. The selection step is shown in FIG.
Is a programmed version of the flowchart shown in
U performs a zigzag scan of the AC coefficient of the quantized DCT coefficient Ruv and performs a first selection step of turning on an embedding flag corresponding to the quantized DCT coefficient having a determination value of 2 or more, if any. This embedding flag is stored in the RAM. When the number C with the embedding flag turned on reaches the embedding amount 16, or when the last quantized DCT
When the zigzag scan is performed up to the coefficient, the first selection step ends. The CPU determines whether the number C of which the flags are turned on is smaller than the embedding amount 16, and if the number is smaller, the second selection step of turning on the embedding flags (16-C) which are turned off in the zigzag scanning order. I do.
The CPU executes the quantization D with the embedding flag turned on.
The CT coefficients are replaced in the above replacement step.

[0070]

According to the embedding apparatus of the present invention, in a quantized DCT coefficient block generated by performing discrete cosine transform (DCT) and quantization on an image, the low-frequency AC coefficient and the absolute value are first. A selection means for selecting a predetermined number of AC coefficients from the AC coefficients having a threshold value or more, and a replacement means for replacing the least significant bit of the predetermined number of AC coefficients selected by the selection means with audio data.

According to this configuration, the embedding apparatus selects an AC coefficient that causes less deterioration of the image when the value of the least significant bit is changed, and embeds the audio data, thereby reducing the deterioration of the image. This has the effect. Here, the deterioration of the image is reduced for the following reason. First, the low-frequency AC coefficient will be described. The low-frequency AC coefficient is characterized in that it is usually quantized at a smaller quantization level than the high-frequency AC coefficient. This is because human visual characteristics are insensitive to high-frequency components and sensitive to low-frequency components, so that high-frequency components are roughly quantized to increase the compression ratio. When decoding, the AC coefficients are inversely quantized at the same value as the quantization level, so that low-frequency AC coefficients are inversely quantized at a small level and high-frequency AC coefficients are inversely quantized at a large level. Becomes Therefore, even if the least significant bit is replaced from 0 to 1, the error after the inverse quantization is smaller at the low frequency than at the high frequency, that is, the image quality is less deteriorated.

An AC coefficient having an absolute value equal to or larger than a threshold value will be described. An AC coefficient having a large absolute value has a smaller rate of change when the least significant bit changes than an AC coefficient having a small absolute value. . For example, comparing the case where the least significant bit of the 16 AC coefficient is 17 and the case where the 0 AC coefficient is 1, the change ratio is smaller when changing from 16 to 17. Also, the AC coefficient having a large absolute value is often at a low frequency (because the low frequency is quantized at a smaller quantization level, so the value is often large), and the inverse quantization is performed accordingly. Later error is small. From these, the AC coefficient whose absolute value is equal to or larger than the threshold value has less image deterioration than the AC coefficient whose absolute value is smaller than the threshold value.

Further, the selecting means includes a first selecting section for selecting an AC coefficient whose absolute value is equal to or greater than a first threshold value, and a selecting means for determining whether the number of AC coefficients selected by the first selecting means is less than a predetermined number. Is less than the first threshold and the lower frequency A
A second selection unit for selecting AC coefficients in order from the C coefficient until a predetermined number is reached. According to this configuration, the embedding apparatus reduces deterioration of image quality when replacement is performed by selecting an AC coefficient having a large absolute value with priority over an AC coefficient of a low frequency. This is A with a large absolute value
This is because the C coefficient causes less deterioration in image quality when the least significant bit is replaced than the low-frequency AC coefficient.

The low-frequency AC coefficient whose absolute value is less than the threshold value reduces an error when the least significant bit is changed due to one factor that the inverse quantization level is small. On the other hand, AC coefficients whose absolute values are equal to or larger than the threshold value are distributed on a relatively low frequency side, and thus have a low inverse quantization level. In addition, the AC coefficient whose absolute value is equal to or larger than the threshold has a smaller change rate when the least significant bit changes than the AC coefficient whose absolute value is smaller than the threshold. As described above, the error of the AC coefficient whose absolute value is equal to or larger than the threshold value is reduced due to two factors, that is, the inverse quantization level is small and the rate of change when the least significant bit changes is small. For this reason, the embedding apparatus converts the AC coefficient having a large absolute value into a low-frequency AC coefficient.
By selecting the coefficient with priority, the deterioration of the image quality is further reduced.

The selecting means further includes a third selecting section for selecting an AC coefficient whose absolute value is equal to or greater than a second threshold value and equal to or greater than a second threshold value, and the replacing means further includes a third selecting section. The audio data is embedded in the second least significant bit of the selected AC coefficient. According to this configuration, the embedding device embeds audio data in the second least significant bit in addition to embedding audio data in the least significant bit, so that there is an effect that more audio data can be embedded. . Further, since the AC coefficient whose absolute value is equal to or larger than the second threshold value, that is, the AC coefficient having a small change rate when the second bit changes from the least significant bit is selected for embedding, the embedding apparatus has less image deterioration. , More voice data can be embedded.

A digital camera according to the present invention is a digital camera in which compressed audio data corresponding to audio data for several seconds is embedded in a compressed image, wherein a quantized DCT coefficient obtained by performing discrete cosine transform and quantization from the compressed image. Acquiring means for obtaining a block, dividing means for dividing the compressed audio data into partial compressed audio data of a predetermined bit, and a low-frequency AC coefficient and an absolute value which are first in a quantized DCT coefficient block to be obtained. From the AC coefficient not less than the threshold value,
There is provided a selecting means for selecting a C coefficient, and a replacing means for replacing a least significant bit of the predetermined number of AC coefficients selected by the selecting means with the partial compressed audio data.

According to this configuration, the digital camera performs the operation of embedding the partially compressed audio data of a predetermined bit in a predetermined number of AC coefficients that requires little image degradation.
Since the processing is performed on the CT coefficient block, a large amount of audio data of predetermined bits × total number of blocks as a whole can be embedded with little deterioration of the image.

[Brief description of the drawings]

FIG. 1 is an external view of a front side of a digital camera 1. FIG.

FIG. 2 is an external view of the back side of the digital camera 1. FIG.

FIG. 3 is a schematic configuration diagram of the digital camera 1.

FIG. 4 is a detailed configuration diagram of FIG. 3;

FIG. 5 is a more detailed configuration diagram of an encoding unit 34.

FIG. 6 is a more detailed configuration diagram of a decoding unit 62.

FIG. 7 shows a relationship between a luminance component Y for one screen and blocks.

FIG. 8A shows a specific example Yxy (x, y = 0 to 7; x, y represents a pixel position in a block) of a pixel of one block of a luminance component Y. (B) D obtained by performing DCT on Yxy
4 shows a CT coefficient block Suv. (C) A specific example of the quantization table Quv is shown. (D) shows a quantized DCT coefficient block Ruv when the DCT coefficient block Suv shown in FIG. 8B is quantized by the quantization table Quv in FIG. 8C.

9 is a detailed configuration diagram of an embedding unit 37. FIG.

FIG. 10 is a flowchart showing a selection process by a selection unit 106;

FIG. 11A shows a quantized DC selected for embedding from the quantized DCT coefficient block Ruv in FIG. 8D.
The T coefficient is indicated by a circle. (B) shows the embedding flag Euv of the selection result when the selection unit 106 performs the selection process on FIG.

FIG. 12A shows a quantized DCT coefficient selected for embedding from a quantized DCT coefficient block different from that shown in FIG. (B) Indicates an embedding flag corresponding to (a). (C) shows an embedded flag of a final selection result by the selection processing of the selection unit 106.

FIG. 13A shows an example of partially compressed audio data. FIG. 12B shows a result when the embedding processing unit 107 embeds the partially compressed audio data in FIG. 12A into the quantized DCT coefficient block in FIG.

FIG. 14 is a detailed configuration diagram of an extraction unit 83.

15 is a detailed configuration diagram of an embedding section 47. FIG.

FIG. 16 shows a quantized DCT in which a selection unit 206 embeds data.
9 is a flowchart illustrating logic for determining the number of coefficients and the bit position to be embedded.

FIG. 17 shows an embedding flag E2uv set according to the result of this determination.

[Explanation of symbols]

 Reference Signs List 3 image encoding unit 35 encoding memory 36 memory card input / output unit 4 audio encoding unit 37 embedding unit 5 audio decoding unit 6 image decoding unit 83 extraction unit 31 imaging unit 33 photographed image memory 34 encoding unit 71 DCT unit 72 quantization unit 74 entropy coding unit 84 entropy decoding unit 82 inverse quantization unit 81 inverse DCT unit 101 compressed image input unit 102 decision value input unit 103 embedding amount input unit 104 compressed audio input unit 105 entropy decoding unit 106 selection Unit 107 embedding processing unit 108 output unit 831 identification unit 832 extraction processing unit

──────────────────────────────────────────────────続 き Continued on the front page (51) Int.Cl. ⁷ Identification symbol FI Theme coat ゛ (Reference) H04N 7/30 H04N 7/133 Z F-term (Reference) 5C022 AA13 AB00 AC01 AC32 AC42 AC71 AC72 5C053 FA07 GB07 GB11 GB22 GB34 GB36 HA33 JA03 JA07 JA12 KA04 KA05 LA01 LA06 5C059 KK02 KK08 LA01 MA00 MA23 MC04 MC14 MC23 MC26 MC32 MC34 PP01 RB06 RC32 RC38 SS15 TA36 TB07 TC04 TC06 TD12 UA02 UA05 UA38 5J064 AA01 BA16 BB11 BC25 BD03

Claims (8)

[Claims] 1. A low-frequency AC coefficient and an AC coefficient whose absolute value is equal to or greater than a first threshold value in a quantized DCT coefficient block generated by performing discrete cosine transform (DCT) and quantization on an image. An embedding apparatus, comprising: a selection unit that selects a predetermined number of AC coefficients from a plurality of AC coefficients; and a replacement unit that replaces least significant bits of the predetermined number of AC coefficients selected by the selection unit with audio data. 2. The method according to claim 1, wherein the selecting unit selects an AC coefficient whose absolute value is equal to or greater than a first threshold value, and the number of AC coefficients selected by the first selecting unit is less than a predetermined number. 2. The embedding apparatus according to claim 1, further comprising: a second selection unit that selects AC coefficients that are less than the first threshold value and that has a predetermined number in order from lower frequency AC coefficients. 3. The method according to claim 1, wherein said selecting means further comprises a signal having an absolute value equal to or greater than a second threshold value larger than said first threshold value.
3. The method according to claim 2, further comprising a third selection unit that selects a C coefficient, wherein the replacement unit embeds audio data in the second least significant bit of the AC coefficient selected by the third selection unit. Embedded device. 4. A quantized DCT coefficient block generated by performing discrete cosine transform (DCT) and quantization on an image, wherein a low-frequency AC coefficient and an AC coefficient whose absolute value is equal to or greater than a first threshold value are included. An embedding method comprising: a selection step of selecting a predetermined number of AC coefficients from the above; and a replacement step of replacing the least significant bit of the predetermined number of AC coefficients selected by the selection step with audio data. 5. The method according to claim 1, wherein the selecting includes selecting an AC coefficient whose absolute value is equal to or greater than a first threshold value, and when the number of AC coefficients selected by the first selecting unit is less than a predetermined number. 5. The embedding method according to claim 4, further comprising the step of: selecting a plurality of AC coefficients from a lower frequency AC coefficient until the number of AC coefficients becomes smaller than a first threshold value. 6. A digital camera that embeds compressed audio data corresponding to audio data for several seconds in a compressed image, wherein the acquisition unit obtains a quantized DCT coefficient block subjected to discrete cosine transform and quantization from the compressed image. Dividing means for dividing the compressed audio data into partial-compressed audio data of a predetermined bit;
Selecting means for selecting the predetermined number of AC coefficients from a coefficient and an AC coefficient whose absolute value is equal to or greater than a first threshold value; and compressing least significant bits of the predetermined number of AC coefficients selected by the selecting means. A digital camera, comprising: replacement means for replacing with audio data. 7. A computer-readable recording medium in which a program for causing a computer to execute a process of embedding audio data in an image is recorded. The program stores the program in a computer by means of discrete cosine transform (DCT) and quantum Selecting a predetermined number of AC coefficients from a low-frequency AC coefficient and an AC coefficient whose absolute value is equal to or greater than a first threshold value in a quantized DCT coefficient block generated by performing Replacing the least significant bit of the predetermined number of AC coefficients selected by the step with audio data. 8. The method according to claim 1, wherein the selecting includes selecting an AC coefficient whose absolute value is greater than or equal to a first threshold value, and when the number of AC coefficients selected by the first selecting unit is less than a predetermined number. 8. The program according to claim 7, wherein the program further comprises: a second selection step of selecting an AC coefficient from a lower frequency AC coefficient until the number of AC coefficients becomes smaller than a predetermined number. Recording medium.