JP3469826B2 - Image compression device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image compression apparatus, and more particularly to an MPEG (Moving Picture Expert Group).
p) Video encoder or JPEG (Joint Photograp)
hic Codibg ExpertGroup) Encoder.

[0002]

2. Description of the Related Art In recent years, the demand for electronic still cameras has increased more and more in place of cameras using silver salt photographic technology since the 19th century.

In an electronic still camera, when transmitting and storing image data, the "JPEG" method is used as a data compression / expansion technique in order to efficiently process the image data by compressing the image data and reducing the data amount. Is used. This JPEG system is based on ISO (International
Organization for Standardization) / IEC (Intarn
ational Electrotechnical Commission) JPE
It is standardized by the G committee (ISO / IEC 10918-1).

The JPEG system is also called a JPEG algorithm, and the core of the technique is the discrete cosine transform (D
CT; Discrete Cosine Transform). And
The JPEG system is not only for electronic still cameras, but also for CD-
It is also widely used for image data processing in ROM (CD-Read Only Memory) systems and the like.

Since the JPEG system can also compress and decompress moving image data, some electronic still cameras using the JPEG system have a moving image shooting function. As described above, a technique for compressing / decompressing moving image data using the JPEG method is M-JPEG (Mo
tion-JPEG).

Information handled by multimedia is
There is a huge amount and variety, and it is necessary to process such information at high speed in order to put multimedia into practical use. In order to process information at high speed, data compression / decompression technology is essential. The "MPEG" method is an example of such data compression / decompression technology. This MPEG system is an MPE under the ISO / IEC umbrella.
It is standardized by the G committee (ISO / IEC JTC1 / SC29 / WG11).

The MPEG system is composed of three parts. Part 1, "MPEG System Part" (ISO /
IEC IS 11172 Part1: Systems) specifies a multiplexing structure and a synchronization method for moving image data (video data) and audio data (audio data). In "MPEG video part" (ISO / IEC IS 11172 Part2: Video) of Part 2, a high-efficiency coding method of moving image data and a format of moving image data are specified. In "MPEG audio part" of Part 3 (ISO / IEC IS 11172 Part3: Audio), a high-efficiency coding method of audio data and an audio data format are specified.

The core of the technology used in the MPEG video part is motion compensation prediction (MC).
ated prediction) and DCT. A coding technique that uses MC and DCT together is called a hybrid coding technique. That is, it can be said that the MPEG system is a technique in which MC is combined with the JPEG system.

The MPEG system is based on various communication such as various storage media (video CD (Compact Disc), CD-ROM, DVD, video tape, memory card using non-volatile semiconductor memory, etc.), LAN (Local Area Network) and the like. Media, various broadcasting media (terrestrial broadcasting, satellite broadcasting,
It is compatible with all transmission media including CATV (Community Antenna Television).

FIG. 9 is a block circuit diagram of a conventional electronic still camera 101 using the JPEG system.

The electronic still camera 101 includes a JPEG core circuit 102, an image pickup device 103, and a signal processing circuit 10.
4, frame buffer 105, display 106, display circuit 107, memory card 108, input / output circuit 10
9 and data buses 110 and 111.

The JPEG core circuit 102 is the DCT circuit 1
21, a quantization circuit 122, a Huffman coding circuit 123,
Code amount counter 124, Huffman decoding circuit 131, inverse quantization circuit 132, inverse DCT (IDCT; Inverse DC)
T) circuit 133, RAM (Random Access Memory) 13
4, 135 and the control core circuit 136.

The control core circuit 136 controls the circuits 102 to 111 of the electronic still camera 101. The image pickup device 103 is composed of a CCD (Charge Coupled Device) or the like, and photographs a subject image to generate an output signal.
The signal processing circuit 104 generates image data for each screen from the output signal of the imaging device 103. The image data for each screen generated by the signal processing circuit 104 is transferred to at least one of the frame buffer 105 and the display circuit 107 via the data bus 110.

The display circuit 107 generates an image signal from the image data for each screen transferred via the data bus 110. The display 106 is a display circuit 107.
The image signal generated by is displayed as a subject image.

The frame buffer 105 is a rewritable semiconductor memory (for example, SDRAM (Synchronous Dynamo)).
mic RAM), DRAM, Rambus DRAM, etc.) and writes and stores the image data for each screen (1 frame) transferred via the data bus 110, and the stored image data for each screen. read out.
The image data for each screen read from the frame buffer 105 is transferred to the DCT circuit 121 of the JPEG core circuit 102 via the data bus 110.

In the JPEG core circuit 102, one screen of image data is divided into a plurality of macroblocks defined by the JPEG standard, and compression / expansion processing is performed for each block.

Here, the DCT circuit 121 and the quantization circuit 1
22. The Huffman encoding circuit 123 constitutes a JPEG encoder, and performs compression processing of image data. Further, the Huffman decoding circuit 131, the inverse quantization circuit 132, and the inverse DCT circuit 133 constitute a JPEG decoder, and perform image data expansion processing.

The DCT circuit 121 includes the frame buffer 1
For each screen of image data read from 05, the image data of one screen is loaded in block units,
Two-dimensional discrete cosine transform (D
CT) to generate DCT coefficients.

The quantizing circuit 122 quantizes the DCT coefficient supplied from the DCT circuit 121 with reference to the quantizing threshold value stored in the quantizing table stored in the RAM 134.

The Huffman coding circuit 123 performs variable length coding on the DCT coefficient quantized by the quantizing circuit 122 by referring to the Huffman code stored in the Huffman table stored in the RAM 135, Compressed image data (hereinafter referred to as compressed image data) is generated one screen at a time.

The code amount counter 124 counts the code amount of the compressed image data for each screen generated by the Huffman coding circuit 123.

The compressed image data generated by the Huffman coding circuit 123 is transferred to at least one of the memory card 108 and the input / output circuit 109 via the data bus 111.

The memory card 108 is the electronic still camera 1.
01 is removably attached, and a flash memory 108a is provided in the memory card 108. The flash memory 108a writes and stores the compressed image data for each screen transferred via the data bus 111, and stores the stored compressed image data in 1
Each screen is read and transferred to the data bus 111.

The input / output circuit 109 receives the compressed image data for each screen transferred via the data bus 111.
The compressed image data is output to an external device (for example, an external display, a personal computer, a printer, etc.) connected to the electronic still camera 101, and the compressed image data input from the external device is transferred to the data bus 111.

The compressed image data read from the memory card 108 or the compressed image data input through the input / output circuit 109 is JPEG through the data bus 111.
It is transferred to the Huffman decoding circuit 131 of the core circuit 102.

The Huffman decoding circuit 131 performs variable-length decoding on the compressed image data for each screen transferred via the data bus 111 with reference to the Huffman code stored in the Huffman table stored in the RAM 135. The converted image data (hereinafter referred to as expanded image data) is generated screen by screen.

The dequantization circuit 132 converts the decompressed image data generated by the Huffman decoding circuit 131 for each screen into R
By performing inverse quantization with reference to the quantization threshold value stored in the quantization table stored in AM134,
Generate DCT coefficients.

The inverse DCT circuit 133 is an inverse quantization circuit 13
A two-dimensional inverse inverse cosine transform (IDCT) is performed on the generated DCT coefficient of 2.

The decompressed image data for each screen, which has been subjected to the inverse discrete cosine transform in the inverse DCT circuit 133, is transferred to the frame buffer 105 via the data bus 110. Then, the frame buffer 105 is transferred from the inverse DCT circuit 133 via the data bus 110.
Image data for each screen is written and stored. The display circuit 107 also generates an image signal from image data for each screen transferred from the frame buffer 105 via the data bus 110, and the image signal is displayed on the display 106 as a subject image.

Here, the quantization threshold value stored in the quantization table stored in the RAM 134 is the compression rate of the compressed image data or the reproduced image obtained by decompressing the compressed image data (display image on the display 106, external device). Image quality of the reproduced image read from the memory card 108 and reproduced, and the reproduced image reproduced by an external device connected to the input / output circuit 109. In addition, RAM135
The Huffman code stored in the Huffman table stored in is a variable-length code that is assigned according to the appearance frequency predicted in advance for the quantized DCT coefficient or decompressed image data, and has a high appearance frequency. Assigned shortly to.

By the way, the maximum code amount of the compressed image data for each screen generated by the Huffman coding circuit 123 is defined according to the image quality mode preset in the electronic still camera 101. This is because the number of screens (number of photos) that can be stored in the memory card 108 is determined by the image quality mode of the electronic still camera 101. For example, the number of photos that can be stored in the memory card 108 is high quality. The number is set to a few in the mode, and a dozen in the low image quality mode. That is, in the high image quality mode, since the compression rate of the compressed image data is set low, the code amount of the compressed image data of one screen becomes large, and the number of photos that can be stored in the memory card 108 becomes small. Further, in the low image quality mode, the compression rate of the compressed image data is set to be high, so that the code amount of the compressed image data of one screen is reduced and the number of photos that can be stored in the memory card 108 is increased.

Therefore, it is necessary to optimize the quantization threshold value and the Huffman code so that the code amount of the compressed image data for each screen generated by the Huffman coding circuit 123 is equal to or less than the specified maximum value. Here, the data amount of the image data for each screen read from the frame buffer 105 (that is, the image data for each screen generated by the signal processing circuit 104) is irrespective of the subject image captured by the imaging device 103. It is constant. Therefore, the compression rate of the compressed image data generated by the Huffman encoding circuit 123 is uniquely calculated with respect to the code amount of the compressed image data, and the compression rate decreases as the code amount increases.
That is, setting the code amount of the compressed image data to be equal to or less than the specified maximum value is nothing more than setting the compression rate of the compressed image data to be equal to or more than the specified minimum value.

However, the compression rate (code amount) of the compressed image data greatly depends on the combination of the quantization threshold and the Huffman code and the subject image. for that reason,
Even if the quantization threshold and Huffman code are the same,
When different subject images are taken, a large difference occurs in the compression rate of the compressed image data. For example, when the subject image is a complex and detailed image such as a large crowd, it is necessary to increase the compression rate and keep the code amount low, and the subject image is a simple image such as a clear sky with no cloud. In this case, it is necessary to reduce the compression rate and increase the code amount. Therefore,
To obtain the compression rate of the compressed image data, each circuit 121-
It is necessary to perform the compression processing by 123 to actually generate the compressed image data, and it is difficult to predict the compression rate before performing the compression processing.

Therefore, the conventional electronic still camera 101
Then, first, the quantization threshold value and the Huffman code are set to prescribed values, the compression processing by each of the circuits 121 to 123 is performed to generate compressed image data, and the code amount is counted by the code amount counter 124. . As a result, if the code amount of the compressed image data is less than the specified maximum value,
Since it can be said that the set quantization threshold and Huffman code are optimized, the quantization threshold and Huffman code are determined for the set values.

Then, the compressed image data created based on the determined quantization threshold value and Huffman code is transferred to the memory card 108 or the input / output circuit 109 via the data bus 111.

However, when the code amount of the compressed image data exceeds the specified maximum value, the quantization threshold value and the Huffman code are reset to values slightly larger than the specified values, and each circuit 121 is reset. The above-mentioned compression processing by 123 to 123 is performed to generate compressed image data again, and the code amount thereof is counted by the code amount counter 124. As a result, when the code amount of the compressed image data becomes less than the maximum value, it can be said that the re-set quantization threshold and Huffman code are optimized. Determine the threshold and Huffman code.

However, if the code amount of the compressed image data still exceeds the maximum value, the quantization threshold and the Huffman code are set to larger values and the compression process is performed again.

As described above, the conventional electronic still camera 10
1, the quantization threshold and the Huffman code are set so that the code amount of the compressed image data is equal to or less than the specified maximum value (that is, the compression rate of the compressed image data is equal to or more than the specified minimum value). To optimize each circuit 121
The above-mentioned compression processing by 123 is repeated. Therefore, it takes some time to determine the quantization threshold and the Huffman code.

[0039]

The conventional electronic still camera 101 has the following problems.

(1) In order to optimize the quantization threshold value and the Huffman code, the compression processing by each of the circuits 121 to 123 is repeatedly performed, so that the imaging device 103 captures only the time required for the compression processing. The time (hereinafter, referred to as a recording waiting time) required for storing the compressed image data in the memory card 108 after performing the process becomes long.

In order to avoid this problem, it is conceivable to set the recording waiting time short beforehand, but in that case, the number of times the circuits 121 to 123 repeatedly perform the compression processing is reduced, The quantization threshold and the Huffman code cannot be optimized, and the quality of the reproduced image obtained by expanding the compressed image data may deteriorate.

(2) Each time the circuits 121 to 123 perform the compression processing, the image data for each screen must be read from the frame buffer 105 and transferred to the DCT circuit 121 via the data bus 110. I have to. That is, each time the circuits 121 to 123 perform the compression processing, the frame buffer 105 must be read-accessed via the data bus 110.

Therefore, access to the frame buffer 105 becomes congested, and the above-mentioned problem causes the recording waiting time to become longer. Further, when the access to the frame buffer 105 is congested, the performance of the access performed from the signal processing circuit 104 and the display circuit 107 to the frame buffer 105 is deteriorated.

To avoid this problem, the bus width of the frame buffer 105 and the data bus 110 may be increased or the speed may be increased.
Since 05 is expensive and consumes a large amount of power, it causes an increase in cost and power consumption of the electronic still camera 101. In particular, since the electronic still camera 101 is driven by a battery, the increase in power consumption is a big problem.

Furthermore, since a camera using the silver salt photographic technique records an image on a film almost immediately after photographing, continuous photographing can be performed one after another. On the other hand, since the electronic still camera 101 requires the recording waiting time, continuous shooting cannot be performed one after another, causing the user to feel stress. In order to eliminate this stress, in recent years, it has been required to shorten the recording waiting time more than ever.

The problems (1) and (2) above are related to JPEG.
This happens not only in the encoder but also in the MPEG encoder. That is, as described above, M
The PEG method is a technology in which MC is combined with the JPEG method, and the MPEG encoder is a JPEG encoder (DC
Since the T circuit 121, the quantization circuit 122, and the Huffman coding circuit 123) are configured by adding an MC circuit,
The problems (1) and (2) of the G encoder also apply to the MPEG encoder.

The present invention has been made in order to solve the above-mentioned problems, and an object thereof is to set the code amount of encoded image data after compression to a predetermined value, It is to realize highly accurate encoding.

[0048]

An image compression apparatus according to a first aspect of the present invention includes an encoder for compressing and encoding input image data and a sample image data processed by the encoder for compression. A control circuit for determining an optimum compression parameter for making the code amount of the image data after compression of the image of one screen to be compressed equal to or less than a predetermined value, and the control circuit includes a plurality of images of the one screen to be compressed. Divide into blocks and extract a predetermined block from the multiple blocks with regularity so that it is evenly extracted over one screen to obtain the sample image data of the first block group.
Configured and configured sample image of the first block group
By performing the compression process on the data, the first pressure
The reduced image data is generated and the first block group
The compression parameter optimized for the sample image data
Calculate the number of blocks more than the first block group.
Of the second block group, which is constructed by regularly extracting
By performing the compression processing on the pull image data,
The second compressed image data is generated and
Of the sample image data of the
The compression parameter is obtained and the number of blocks in the first block group is calculated.
Converted when the number of blocks in the second block group is the same
Of the first compressed image data and the second compressed image data
The compression error, which is the difference in code amount, is calculated, and the compression error is allowed.
If the value is within the acceptable range, the image data of the first block group
The optimized compression parameters are
The point is to decide as a meter .

Image compression apparatus according to the first aspect
Is the first block if the compression error is within the allowable value.
To the compression parameters optimized for the group of image data
Instead, it is optimized for the image data of the second block group.
The compressed parameter as the optimum compression parameter
You may decide.

In the image compression apparatus according to the first and second aspects, the control circuit extracts a predetermined block from the plurality of blocks at a predetermined interval to generate the sample image data, or the plurality of blocks. It is preferable that the sample image data is generated by extracting a block group adjacent to each other from among the blocks at a predetermined interval.

The image compression apparatus according to the second aspect of the present invention includes an encoder that compresses and encodes input image data, and a code amount that counts the code amount of the compressed image data generated by this encoder. A counter,
The compression amount by the encoder is repeated while changing the compression parameter of the encoder until the code amount of the compressed image data counted by the code amount counter becomes equal to or less than a predetermined value, thereby making the code amount of the compressed image data a predetermined value. And a control circuit for determining the optimum compression parameter for dividing the image of one screen to be compressed into a plurality of blocks, and selecting a predetermined block from the plurality of blocks into one screen. The sample of the first block group is extracted with regularity so that it is extracted evenly over
The image data of the first block
To perform the compression process on sample image data
The first compressed image data is generated from the first compressed image data and
Before optimized for sample image data of lock group
The compression parameters are calculated and are larger than those in the first block group.
A second block constructed by regularly extracting a number of blocks
The compression processing is performed on the sample image data of the
By generating the second compressed image data by
Optimization for the sample image data of the second block group
The compressed parameters are calculated, and the block of the first block group is calculated.
The number of locks and the number of blocks in the second block group are the same
The first compressed image data and the second compressed image that are sometimes converted
Calculating a compression error is the difference between the code amount of the image data, the pressure
If the reduction error is within the allowable value, the image data of the first block group
The compression parameters optimized for the data
The main point is to decide as a compression parameter .

Image compression apparatus according to the second aspect
Is the first block if the compression error is within the allowable value.
To the compression parameters optimized for the group of image data
Instead, it is optimized for the image data of the second block group.
The compressed parameter as the optimum compression parameter
You may decide.

Further, the encoder performs discrete cosine transform on the image data forming one screen image,
A discrete cosine transform circuit for generating a discrete cosine transform coefficient, and a discrete cosine transform coefficient supplied from the discrete cosine transform circuit are quantized with reference to a quantization threshold stored in a preset quantization table. And a discrete cosine transform coefficient quantized by the quantizing circuit, by performing variable length coding with reference to a Huffman code stored in a preset Huffman table, compressed image data And a Huffman encoding circuit for generating.

Further, it is desirable that the compression parameter comprises at least one of the quantization threshold and the Huffman code.

In the embodiments of the invention described below, the "sample image data" described in the claims or the means for solving the problems are screen P2 and screen P3.
And the “first block group” and the “second block group” are
It corresponds to the block Bs1 forming the screen P2 and the block Bs2 forming the screen P3, or the block Bs2 forming the screen P3 and the block Bs3 forming the screen P4.

[0056]

DETAILED DESCRIPTION OF THE INVENTION Embodiments embodying the present invention will be described below with reference to the drawings. In the present embodiment, the same components as those in the conventional form are designated by the same reference numerals, and detailed description thereof will be omitted.

FIG. 1 shows a block circuit diagram of an electronic still camera 1 of this embodiment using the JPEG system.

In the conventional electronic still camera 101,
When determining the quantization threshold and the Huffman code, the image data transferred through the path of frame buffer 105 → data bus 110 → DCT circuit 121 is the image data for one screen.

On the other hand, in the electronic still camera of this embodiment, when determining the quantization threshold and the Huffman code, one screen of the image data read from the frame buffer 105 is divided into a plurality of blocks, and the blocks are divided. Of the image data of a predetermined block among them, the frame buffer 105 → data bus 110 → DCT circuit 121
I am trying to be transferred by the route.

That is, as shown in FIG. 2A, one screen P1 of the image data read from the frame buffer 105 is arranged in 16 rows and 16 columns (16 rows × 16 columns), 256 pixels each.
It is divided into individual blocks Ba. The block Ba is set in units of macroblocks, and one block Ba is set.
Is composed of m vertical blocks and n horizontal blocks (m rows × n rows, where m and n are natural numbers).

Next, one screen P is selected from each block Ba.
A block Ba (hereinafter, referred to as “Bs1” to be distinguished from each other) arranged every three blocks in the vertical and horizontal directions from the upper left end of 1 is selected, and as shown in FIG. 4 lines ×
A screen P2 including 16 blocks Bs1 (4 columns) is created.

Then, only the image data for the screen P2 is read from the frame buffer 105, and only the image data for the screen P2 is transferred through the path of the frame buffer 105 → data bus 110 → DCT circuit 121, and the quantization threshold value is transferred. The Huffman code is set to a preset value for the screen P2, the compression processing is performed by the circuits 121 to 123 to generate compressed image data, and the code amount is counted by the code amount counter 124. To do. By repeating the compression processing by each of the circuits 121 to 123 until the code amount of the compressed image data becomes equal to or less than the maximum value specified for the screen P2, the quantization threshold value based on the image data of the screen P2. And the Huffman code is provisionally determined.

Then, as shown in FIG. 3A, the screen P
From among the blocks Ba of No. 1, every other block Ba (hereinafter referred to as “Bs2” to be distinguished from each other) arranged in the vertical and horizontal directions from the upper left end of the screen P1 is selected, and FIG. As shown in (b), a screen P3 including 64 blocks Bs2 (8 rows × 8 columns) is created.

Then, only the image data for the screen P3 is read out from the frame buffer 105, and only the image data for the screen P3 is transferred through the path of the frame buffer 105 → data bus 110 → DCT circuit 121 to obtain the screen P2.
With reference to the quantization threshold value and Huffman code temporarily determined based on the image data, the circuits 121 to 123 perform the compression process to generate compressed image data, and the code amount is calculated by the code amount counter 124. To count. By repeating the compression process by each of the circuits 121 to 123 until the code amount of the compressed image data becomes equal to or less than the maximum value specified for the screen P3, the quantization threshold value based on the image data of the screen P3. And the Huffman code is provisionally determined.

Here, the code amounts of the image data of the screens P1 and P2 have the relationship shown in the equation (1).

SP1 = SP2 × 16 + E1 (Equation 1) where SP1; code amount SP2 of image data of screen P1; code amount E1 of image data of screen P2; compression between image data of screens P1 and P2 The error, that is, the screen P2 is composed of 16 blocks Bs1 (= Ba) evenly arranged on the screen P1 composed of 256 blocks Ba. Therefore, the size of the screen P2 is 1/16 of the screen P1. Become. However, since the information of the subject image existing in the portion excluding the blocks Bs1 configuring the screen P2 from the screen P1 is not included in the image data for the screen P2, the code amount SP2 of the image data of the screen P2 is The value simply multiplied by 16 does not become the same as the code amount SP1 of the image data of the screen P1. Therefore, the value obtained by adding the compression amount E1 resulting from the information of the subject image existing in the portion excluding the blocks Bs1 forming the screen P2 from the screen P1 to the value obtained by multiplying the code amount SP2 by 16 is the code amount SP1. It becomes a value.

At this time, as described above, the screen P2 is created by selecting, from each block Ba, blocks Bs1 which are arranged every three rows in the vertical and horizontal directions from the upper left end of the screen P1. That is, since the blocks Bs1 that are evenly arranged on the screen P1 are extracted (because the blocks Bs1 are extracted from the 256 blocks Ba at equal intervals), the characteristics of the screen can be extracted evenly. Therefore, the compression error E1 can be suppressed to be small regardless of the information of the subject image.

For example, in the case where a complicated and detailed image such as a large crowd and a simple image such as a clear sky without clouds are mixed in one screen, the block Bs1 is replaced by the block Bs1. In the case of random extraction from the inside, one of fine image data and simple image data is likely to be biased and the compression error E1 becomes large, but in the present embodiment, such a problem is unlikely to occur.

Here, the relationship between the blocks Bs1 and Bs2 forming the screens P2 and P3 is as shown in FIG. That is, as shown in FIG. 4A, the screen P3
Of each of the 64 blocks Bs2, the block Bs2 arranged every other row in the vertical and horizontal directions from the upper left end of the screen P3 is selected as the block Bs1, and as shown in FIG. The screen P2 is composed of the individual blocks Bs1.

Therefore, the code amount of the image data on each of the screens P2 and P3 has the relationship shown in the equation (2).

SP3 = SP2 × 4 + E2 (Equation 2) where SP3: code amount E2 of image data of screen P3; compression error between image data of screens P2 and P3, that is, 64 blocks Bs2 Since the screen P2 is composed of 16 blocks Bs1 evenly arranged on the screen P3, the size of the screen P2 is the size of the screen P2.
It becomes 1/4 of 3. However, since the information of the subject image existing in the portion excluding each block Bs1 forming the screen P2 from the screen P3 is not included in the image data of the screen P2, the code amount SP2 of the image data of the screen P2 is The value simply multiplied by 4 does not become the same as the code amount SP3 of the image data of the screen P3. Therefore, the code amount SP2 is multiplied by 4,
The value of the code amount SP3 is a value obtained by adding the compression error E2 caused by the information of the subject image existing in the portion excluding the blocks Bs1 configuring the screen P2 from the screen P3.

In the present embodiment, if the compression error E2 is within the allowable value, it is considered that the compression error E1 is also within the allowable value, and the quantization threshold temporarily determined based on the image data of the screen P2. And the Huffman code,
This determination is made assuming that the quantization threshold and the Huffman code are optimized for the image data of the screen P1.

At this time, as described above, the screen P3 is created by selecting, from among the blocks Ba, the blocks Bs1 which are arranged every other row in the vertical and horizontal directions from the upper left end of the screen P1. That is, since the blocks Bs1 that are evenly arranged on the screen P1 are extracted, the characteristics of the screen can be taken out evenly, and like the above-described compression error E1, what information about the subject image is. However, the compression error E2 can be suppressed to be small.

Then, all the image data of the screen P1 is read from the frame buffer 105, and the image data of the screen P1 is transferred from the frame buffer 105 to the data bus 11
The data is transferred through the path of the 0 → DCT circuit 121, and the compression processing is performed by each of the circuits 121 to 123 with reference to the quantization threshold and the Huffman code which have been finally determined, and compressed image data is generated. The compressed image data is the data bus 1
Via the memory card 108 or the input / output circuit 10
9 is transferred.

When the compression error E2 exceeds the allowable value, each block B of the screen P1 is displayed as shown in FIG. 5 (a).
From among a, select one of the blocks Ba (hereinafter referred to as “Bs3” to be distinguished) arranged one after another in the vertical and horizontal directions from the upper left end of the screen P1 and arranged in succession. As shown in FIG. 5B, the (12 rows × 12 rows) 144
A screen P4 composed of individual blocks Bs3 is created.

Then, only the image data for the screen P4 is read from the frame buffer 105, and only the image data for the screen P4 is transferred through the path of the frame buffer 105 → data bus 110 → DCT circuit 121, and the screen P3 is displayed.
With reference to the quantization threshold value and Huffman code temporarily determined based on the image data, the circuits 121 to 123 perform the compression process to generate compressed image data, and the code amount is calculated by the code amount counter 124. To count. By repeating the compression processing by each of the circuits 121 to 123 until the code amount of the compressed image data becomes equal to or less than the maximum value defined for the screen P4, the quantization threshold value based on the image data of the screen P4. And the Huffman code is provisionally determined.

Here, the relationship between the blocks Bs2 and Bs3 forming the screens P3 and P4 is as shown in FIG. That is, as shown in FIG. 6A, the screen P4
6 out of the 144 blocks Bs3, the block Bs3, which is arranged two consecutively in the vertical and horizontal directions from the upper left end of the screen P4, is selected as the block Bs2.
As shown in (b), the screen P3 is composed of the 64 blocks Bs1.

Therefore, the code amount of the image data on each of the screens P3 and P4 has the relationship shown in the equation (3).

SP4 = SP3 × 144/64 + E3 (Equation 3) where SP4: Code amount E3 of image data of screen P3; compression error between image data of screens P3 and P4, that is, 144 blocks Bs3 Since the screen P3 is composed of 64 blocks Bs2 evenly arranged on the screen P4, the size of the screen P3 is 64/144 of the screen P4. However, since the information of the subject image existing in the portion excluding the blocks Bs2 constituting the screen P3 from the screen P4 is not included in the image data of the screen P3, the code amount SP3 of the image data of the screen P3 is included.
The value simply multiplied by 144/64 does not become the same as the code amount SP4 of the image data of the screen P4. Therefore, the code amount S
A value obtained by adding a compression error E3 resulting from the information of the subject image existing in the part of the screen P4 excluding the blocks Bs2 constituting the screen P3 to the value obtained by multiplying P3 by 144/64 is the value of the code amount SP4. Become.

In this embodiment, when the compression error E3 is within the allowable value, it is considered that the compression error E1 is also within the allowable value, and the quantization threshold value and the Huffman code provisionally determined based on the screen P3. Is determined as the Huffman code and the quantization threshold optimized for the image data of the screen P1.

At this time, as described above, the screen P4 is a group of adjacent blocks in which three blocks are arranged consecutively after one block Ba is placed from the upper left end of the screen P1 in the vertical and horizontal directions. It is created by selecting Bs3. That is, since the blocks Bs3 that are evenly arranged on the screen P1 are extracted (because the block groups Bs3 that are adjacent to each other at equal intervals are extracted from the 256 blocks Ba), the characteristics of the screen are averaged. Can be taken out to
Similar to the compression errors E1 and E2 described above, the compression error E3 can be suppressed to be small regardless of the information of the subject image.

Then, all the image data of the screen P1 is read from the frame buffer 105, and the image data of the screen P1 is transferred from the frame buffer 105 to the data bus 11
The data is transferred through the path of the 0 → DCT circuit 121, and the compression processing is performed by each of the circuits 121 to 123 with reference to the quantization threshold and the Huffman code which have been finally determined, and compressed image data is generated. The compressed image data is the data bus 1
Via the memory card 108 or the input / output circuit 10
9 is transferred.

When the compression error E3 exceeds the allowable value, the screen P1 is displayed as in the conventional electronic still camera 101.
The quantization threshold value and the Huffman code are determined based on all the image data of 1, and the compressed image data is generated by the determined quantization threshold value and the Huffman code and transferred to the data bus 111.

FIG. 7 is a flow chart for explaining the operation of this embodiment.

First, in step (hereinafter referred to as "ST") 1, the image data of screen P2 is displayed on screen P2.
Compressed image data is generated by referring to the quantization threshold value and the Huffman code set for, and the code amount SP2 is obtained, and the quantization threshold value and the Huffman code based on the image data of the screen P2 are provisionally determined. To do.

Next, in ST2, for the image data of the screen P3, compressed image data is generated by referring to the Huffman code and the quantization threshold provisionally determined based on the image data of the screen P2, and the code amount SP3 thereof is set. At the same time, the quantization threshold and the Huffman code based on the image data of the screen P3 are provisionally determined.

Next, in ST3, the compression error E2 is calculated from the respective code amounts SP2 and SP3. If the compression error E2 is within the allowable value (ST3: YES), the process proceeds to ST4, and if it exceeds the allowable value (ST3 : NO) moves to ST5.

When the process shifts from ST3 to ST4, ST4
Then, the quantization threshold value and the Huffman code temporarily determined based on the image data of the screen P2 are finally determined as the quantization threshold value and the Huffman code optimized for the image data of the screen P1. Transition to ST6.

In ST6, the compressed image data is generated for the image data of the screen P1 based on the quantization threshold value and the Huffman code finally determined in ST4, and the compressed image data is transferred to the data bus 111 and encoded. To finish.

Further, in ST5, with respect to the image data of the screen P4, compressed image data is generated by referring to the Huffman code and the quantization threshold provisionally determined based on the image data of the screen P3, and the code amount SP4 is obtained. At the same time, the quantization threshold and the Huffman code based on the image data of the screen P4 are provisionally determined.

Next, in ST7, the compression error E3 is calculated from each of the code amounts SP3 and SP4. If the compression error E3 is within the allowable value (ST7: YES), the process proceeds to ST4, and if it exceeds the allowable value (ST7: : NO) moves to ST8.

When shifting from ST7 to ST4, ST4
Then, the quantization threshold and Huffman code provisionally determined based on the image data of the screen P3 are finally determined as the quantization threshold and Huffman code optimized for the image data of the screen P1.

Further, in ST8, as with the conventional electronic still camera 101, with respect to the image data of the screen P1, the quantization threshold and the Huffman code are determined based on all the image data of the screen P1, and then to ST6. Transition.

In ST6, the compressed image data is generated for the image data of the screen P1 based on the quantization threshold and the Huffman code determined in ST8, and the compressed image data is transferred to the data bus 111 to be encoded. finish.

The processing of ST1 to ST8 is performed by the control core circuit 2 controlling each of the circuits 102 to 111. That is, the control core circuit 2 divides the image of the screen P1 into a plurality of blocks Ba, selects a predetermined block Ba while changing the number of blocks to be selected from the blocks Ba, and the image of the selected block Ba. By repeating the compression process on the data, an operation is performed to determine the optimum quantization threshold value and the parameter including the Huffman code for making the code amount of the compressed image data equal to or less than the specified maximum value. That is, the control core circuit 2 is a RISC (Reduced Instruction).
Set Computer) -provides a CPU function, and has a control core circuit 2
The RISC-CPU processes the above operation by software.

As described above, according to this embodiment, the following actions and effects can be obtained.

(1) The recording waiting time can be shortened.

That is, in ST1, ST2 and ST5 described above, the screens P2 and P are read from the frame buffer 105, respectively.
Image data for 3 and P4 are read out, and the image data are transferred to the DCT circuit 121 via the data bus 110. The screen size of each screen P2, P3, P4 is smaller than the screen size of the screen P1, and the data amount of image data is proportional to the screen size. Therefore, the image data amount of the screen P2 is 1/16, the image data amount of the screen P3 is 1/4, and the image data amount of the screen P4 is 144/256 with respect to the image data amount of the screen P1.

The time required for the compression processing by each of the circuits 121 to 123 is proportional to the amount of image data. Therefore, the processing time of ST1, ST2, ST5 is shorter than the processing time of ST8. For a normal subject image, ST1 → ST2 → ST3 → ST
The processing is performed in the order of 4 → ST6, rarely transits to ST5, and extremely rarely transits to ST8. Therefore, in the present embodiment, the time required for final determination of the quantization threshold and Huffman coding is shorter than that of the conventional electronic still camera 101.

Further, by performing the processing of ST1 to ST8, the recording waiting time is shortened, and each circuit 121
It is possible to sufficiently increase the number of times of repeating the compression processing by 123 to 123. Therefore, the quantization threshold value and the Huffman code can be optimized, and highly accurate compressed image data can be generated. Therefore, it is possible to prevent deterioration in image quality of a reproduced image obtained by expanding the compressed image data.

(2) Since the image data amount of each screen P2, P3, P4 is smaller than the image data amount of the screen P1, S
Frame buffer 105 at T1, ST2 and ST5
The time required for the read access of the frame buffer 105 is shorter than the time required for the read access of the frame buffer 105 in ST8. Therefore, since access to the frame buffer 105 is reduced, the frame buffer 105
Also, it is possible to avoid the congestion of access to the frame buffer 105 without increasing the bus width of the data bus 110.

Therefore, the recording waiting time can be further shortened in combination with the effect (1). Further, it is possible to improve the performance of the access made from the signal processing circuit 104 and the display circuit 107 to the frame buffer 105.

(3) The screens P2, P3 and P4 are the same as the screen P.
Since the blocks Bs1 (block group Bs3 in the case of the screen P4) evenly arranged in 1 are extracted, the characteristics of the screen can be taken out evenly, and the information of the subject image is what. Even if each compression error E1, E2, E3
Can be kept small. As a result, highly accurate encoding can be realized.

The present invention is not limited to the above embodiment, but may be modified as follows, and even in that case, the same or higher actions and effects can be obtained.

(A) In the case of shifting from ST3 to ST4,
In ST4, the quantization threshold value and the Huffman code temporarily determined based on the image data of the screen P3 instead of the screen P2 are regarded as the quantization threshold value and the Huffman code optimized for the image data of the screen P1. Book decision.

Then, in the case of shifting from ST7 to ST4, in ST4, the quantization threshold value and the Huffman code temporarily determined based on the image data of the screen P4, not the screen P3, are optimized for the image data of the screen P1. The quantization threshold and the Huffman code that have been converted into the Huffman code are determined.

(B) In the above embodiment, the screen P1 is changed to (1
Although the screen P1 is divided into 256 blocks Ba (6 rows × 16 rows), the screen P1 may be divided into r vertical blocks and s horizontal blocks (r rows × s rows, where r and s are natural numbers). .
In this case also, the block Ba needs to be set in units of macroblocks.

(C) In the above embodiment, the screen P2 is displayed as (4
16 blocks Ba (row x 4 rows), and screen P3
Is composed of 64 blocks Ba of (8 rows × 8 rows), and the screen P4 is composed of 144 blocks Ba of (12 rows × 12 rows).
However, the number of blocks Ba forming each of the screens P2 to P4 may be set to a value other than this.

For example, the screen P2 is composed of (8 rows × 8 rows) and the screen P3 is composed of (12 rows × 12 rows) blocks Ba, and the screen P4 is omitted. In this case, ST5, ST7
The process of is also omitted, and if ST3: NO, the process proceeds to ST8.

Further, the screen P2 is (2 rows × 2 rows), the screen P
3 is composed of (4 rows × 4 rows), screen P4 is composed of (8 rows × 8 rows) of block Ba, and screen P5 composed of (12 rows × 12 rows) of 144 blocks Ba is added. To do.
In this case, if the compression error E3 exceeds the allowable value in ST7, the code amount SP5 of the image data of the screen P5 is obtained, and the quantization threshold and the Huffman code are provisionally determined based on the image data of the screen P5. Then, the compression error E4 between the image data of the screens P4 and P5 is calculated. If the compression error E4 is within the allowable value, the process proceeds to ST4, and if it exceeds the allowable value, the process proceeds to ST8.

Further, the numbers of the vertical and horizontal directions of the blocks Ba forming the screens P2 to P4 may be different, and for example, the screen P2 (6 lines × 4 lines) and the screen P3 (8 lines × 10).
Line) and the screen P4 may be configured by (14 lines × 12 lines) of blocks Ba.

(D) As shown in FIG. 8, the screen P1 is divided into nine areas Bb of 3 vertically and horizontally (3 rows × 3 columns), and each area Bb is 16 vertically and horizontally (16 rows × 16 rows × 3 columns). 2 in 16 columns)
It is divided into 56 blocks Ba. Then, in the same manner as in the above embodiment, ST1 to ST5, ST for each region Bb
By performing the same processing as 7 and ST8, each area Bb
The quantization threshold and Huffman code are determined for each.

In this case, the processing of ST5 is performed only on the area Bb where the compression error E2 exceeds the allowable value, and the processing of ST8 is performed only on the area Bb where the compression error E3 exceeds the allowable value. Therefore, when there are many areas Bb in which the compression error E2 is within the allowable value or when there are many areas Bb in which the compression error E3 is within the allowable value, the screen P1 is displayed as in the above embodiment.
The time required for the final determination of the quantization threshold value and the Huffman code is shorter than that in the case where is not divided into regions Bb, and the effect of the above-described embodiment can be further enhanced.

The screen P1 is not limited to be divided into nine areas Bb of (3 rows × 3 columns), and the screen P1 may be vertically p and horizontally q.
You may make it divide | segment into the area | region Bb of piece (p row x q row. P and q are natural numbers).

(E) Not limited to electronic still cameras, CD-
It may be applied to an image compression device for image data such as a ROM system.

(F) MP is not limited to the JPEG encoder
It may be applied to an EG encoder. As mentioned above, M
The PEG method is a technology in which MC is combined with the JPEG method, and the MPEG encoder is a JPEG encoder (DC
An MC circuit is added to the T circuit 121, the quantization circuit 122, and the Huffman coding circuit 123).

(G) Each circuit 121 to 123, 131 to 1
The signal processing in 33 is replaced with software-like signal processing using a CPU.

[0118]

In the image compression apparatus of the present invention, the compression process is repeated for the image data of the selected block to determine the optimum compression parameter, so that the compression is performed for all the image data of one screen. As compared with the case of repeating the process, it is possible to shorten the time required to repeat the compression process for determining the optimum compression parameter, and thus it is possible to realize high-speed and high-precision encoding of image data. it can.

Furthermore, since the compression error in determining the optimum compression parameter is suppressed to a small value, it is possible to realize a higher accuracy.

[Brief description of drawings]

FIG. 1 is a block circuit diagram of an electronic still camera according to an embodiment of the present invention.

FIG. 2 is a schematic diagram for explaining the operation of an embodiment embodying the present invention.

FIG. 3 is a schematic diagram for explaining the operation of the present embodiment.

FIG. 4 is a schematic diagram for explaining the operation of the present embodiment.

FIG. 5 is a schematic diagram for explaining the operation of the present embodiment.

FIG. 6 is a schematic diagram for explaining the operation of the present embodiment.

FIG. 7 is a flowchart for explaining the operation of this embodiment.

FIG. 8 is a schematic diagram for explaining an operation of a modified example of this embodiment.

FIG. 9 is a block circuit diagram of a conventional electronic still camera.

[Explanation of symbols]

1 ... Electronic still camera 2 ... Control core circuit 102 ... JPEG core circuit 105 ... Frame buffer 110, 111 ... Data bus 121 ... DCT circuit 122 ... Quantization circuit 123 ... Huffman coding circuit 124 ... Code amount counter 131 ... Huffman decoding circuit 132 ... Inverse quantization circuit 133 ... Inverse DCT circuit P1, P2, P3, P4 ... screen Ba, Bs1, Bs2, Bs3 ... Block Bb ... area

Front page continuation (58) Fields surveyed (Int.Cl. ⁷ , DB name) H04N ^7/ 24-7/68 H04N 1/41-1/419

Claims (6)

(57) [Claims] 1. An encoder that compresses and encodes input image data, and a sample image data that is processed by the encoder to process a code amount of compressed image data of one screen image to be compressed by a predetermined value. And a control circuit for determining an optimum compression parameter for dividing the image of one screen to be compressed into a plurality of blocks, and selecting a predetermined block from the plurality of blocks. Sample image data of the first block group extracted with regularity so that it is extracted evenly across the screen
To the sample image data of the configured first block group.
On the other hand, by performing the compression process, the first compressed image data
Data and generating a sample of the first block group
The compression parameters optimized for the image data
And regularly extract more blocks than the first block group.
Sample image data of the second block group constructed by outputting
To the second compressed image
Generate the data and sump the second block group
Compression parameter optimized for image data
The determined number of blocks of the first block group and the block of the second block group
The first compressed image converted when the same number
A compression error that is the difference in the code amount between the data and the second compressed image data
If the difference is calculated and the compression error is within the allowable value, the first block group
The compression parameters optimized for the image data are
An image compression device characterized in that it is determined as an optimum compression parameter . 2. An encoder that compresses and encodes input image data, and a sample image data is processed by this encoder so that a code amount of the compressed image data of one screen image to be compressed is a predetermined value. And a control circuit for determining an optimum compression parameter for dividing the image of one screen to be compressed into a plurality of blocks, and selecting a predetermined block from the plurality of blocks. Sample image data of the first block group extracted with regularity so that it is extracted evenly across the screen
To the sample image data of the configured first block group.
On the other hand, by performing the compression process , the first compressed image data
Data and generating a sample of the first block group
The compression parameters optimized for the image data
And regularly extract more blocks than the first block group.
Sample image data of the second block group constructed by outputting
To the second compressed image
Generate the data and sump the second block group
Compression parameter optimized for image data
The determined number of blocks of the first block group and the block of the second block group
The first compressed image converted when the same number
A compression error that is the difference in the code amount between the data and the second compressed image data
If the difference is calculated and the compression error is within the allowable value, the second block group
The compression parameters optimized for the image data are
An image compression device characterized in that it is determined as an optimum compression parameter . 3. An encoder for compressing and encoding input image data, a code amount counter for counting the code amount of the compressed image data generated by this encoder, and compressed image data counted by this code amount counter. Until the code amount of is less than or equal to a predetermined value, by repeating the compression process by the encoder while changing the compression parameter of the encoder, the optimum compression parameter for making the code amount of the compressed image data less than or equal to the predetermined value And a control circuit for determining, wherein the control circuit divides an image of one screen to be compressed into a plurality of blocks, and a predetermined block is uniformly extracted from the plurality of blocks over one screen. The sample image data of the first block group is constructed by extracting with regularity.
Forms, with respect to sample image data of the first block group configured
By performing the compression process, the first compressed image data
And a sample image of the first block group
Find the compression parameters optimized for the data
Therefore, a larger number of blocks than the first block group is regularly extracted.
Sample image data of the second block group constructed by outputting
To the second compressed image
Generate the data and sump the second block group
Compression parameter optimized for image data
The determined number of blocks of the first block group and the block of the second block group
The first compressed image and click number is converted to come to have the same
A compression error that is the difference in the code amount between the data and the second compressed image data
If the difference is calculated and the compression error is within the allowable value, the first block group
The compression parameters optimized for the image data are
An image compression device characterized by being determined as an optimum compression parameter . 4. An encoder for compressing and encoding input image data, a code amount counter for counting the code amount of the compressed image data generated by this encoder, and compressed image data counted by this code amount counter. Until the code amount of is less than or equal to a predetermined value, by repeating the compression process by the encoder while changing the compression parameter of the encoder, the optimum compression parameter for making the code amount of the compressed image data less than or equal to the predetermined value And a control circuit for determining, wherein the control circuit divides an image of one screen to be compressed into a plurality of blocks, and a predetermined block is uniformly extracted from the plurality of blocks over one screen. The sample image data of the first block group is constructed by extracting with regularity.
Forms, with respect to sample image data of the first block group configured
By performing the compression process, the first compressed image data
And a sample image of the first block group
Find the compression parameters optimized for the data
Therefore, a larger number of blocks than the first block group is regularly extracted.
Sample image data of the second block group constructed by outputting
To the second compressed image
Generate the data and sump the second block group
Compression parameter optimized for image data
The determined number of blocks of the first block group and the block of the second block group
The first compressed image converted when the same number
A compression error that is the difference in the code amount between the data and the second compressed image data
If the difference is calculated and the compression error is within the allowable value, the second block group
The compression parameters optimized for the image data are
An image compression device characterized in that it is determined as an optimum compression parameter . 5. The discrete cosine transform circuit for performing a discrete cosine transform on image data forming one screen image to generate a discrete cosine transform coefficient,
A quantization circuit that quantizes the discrete cosine transform coefficient supplied from the discrete cosine transform circuit with reference to a quantization threshold value stored in a preset quantization table; A Huffman coding circuit that generates compressed image data by performing variable length coding on the converted discrete cosine transform coefficient with reference to a Huffman code stored in a preset Huffman table. The image compression apparatus according to any one of claims 1 to 4, which is characterized. 6. The image compression apparatus according to claim 5, wherein the compression parameter includes at least one of the quantization threshold and Huffman code.