JP3230336B2 - Recompression device and recompression system - Google Patents

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION The present invention relates to a direct
By performing cross transformation, quantization, and Huffman coding,
In devices and systems that can compress images,
The present invention relates to a technique capable of efficiently performing compression .

[0002]

2. Description of the Related Art Image information is compressed so that it can be transmitted or stored with a small amount of transmission or storage. A technique for compressing a still image is well known, and for example, JPEG (Joint)
(Photograph Expert Group) method. In the JPEG system, one screen image is 8 pixels x
The image data is divided into blocks of 8 pixels, converted into a frequency domain by a two-dimensional discrete cosine transform (DCT) on a block basis, the transform coefficients are quantized, and Huffman coding is performed.

[0003] The amount of data after compression strongly depends on the spatial frequency characteristics of an image and the above-mentioned quantization coefficient. Therefore, even if the quantization coefficients are the same, the data amount after compression differs depending on the image to be compressed. When storing compressed data in a storage device or transmitting the compressed data via a transmission line, there is a demand that the amount of compressed data should be suppressed to or less than a predetermined value.

In order to respond to this, in the conventional example, when the amount of compressed data exceeds a predetermined value, the compressed data is temporarily expanded and then compressed with a different quantization coefficient (ie, DC
T transformation, quantization and Huffman coding)
This was repeated until the desired amount of compressed data was reached.

[0005]

However, when compression is performed, D
The CT transform and the inverse transform at the time of decompression are two-dimensional matrix operations, and it takes a long time even with high-speed hardware. Also,
Since the image restored by the decompression is compressed again, the image quality deteriorates if the calculation accuracy is not sufficient.

The present invention has been made in view of the above conventional example.
And orthogonally transform, quantize, and Huffman code the image
A device capable of executing an encoding algorithm such as JPEG encoding.
Is different from the quantization table used for encoding.
To obtain the compression result encoded using the quantization table
The goal is to increase the processing efficiency when recompressing
Target. Further, the present invention provides the above encoding algorithm
A series of codes in the compression unit assuming that they are applied
That recompression can be performed without affecting the
For further purposes.

[0007]

According to the present invention, there is provided a recompressing apparatus comprising: a transforming means for orthogonally transforming image data;
The transform coefficient obtained by the transformation is calculated according to the first quantization table.
Quantizing means for quantizing, and an output obtained by the quantization
First Huffman encoding means for Huffman encoding;
The compressed image data obtained by Huffman coding
Huffman decoding means for decoding, and Huffman decoding
And the quantization output according to the first quantization table obtained in
Output to the output quantized according to the second quantization table
Changing means for changing, obtained by quantization of said changing means
Second Huffman encoding means for Huffman encoding the output.
Characterized in that it comprises.

[0008] In the recompression system according to the present invention , the encoding unit
And a recompression system comprising a recompression unit, wherein the code
Converting means for orthogonally transforming the image data;
The transform coefficients obtained by the cross transform are calculated according to the first quantization table.
First quantizing means for quantizing by
Huffman encoding means for Huffman encoding the output
Wherein the recompression unit obtains the data obtained by the Huffman coding.
Decoding Huffman decoding of compressed image data
And the first unit obtained by the Huffman decoding.
The quantization output according to the quantization table is converted to a second quantization table.
Change means for changing to an output quantized according to the
Huffman coding of the output obtained by the quantization of the
And second Huffman encoding means.
You.

[0009]

[0010]

Embodiments of the present invention will be described below with reference to the drawings.

FIG. 1 is a block diagram showing a schematic configuration of an embodiment of the present invention. 10 is a CPU for controlling the whole, 12 is an image scanner for reading an image, 14 is a printer, 1
6 is an image display device such as a CRT, 18 is a compression circuit for compressing image data, 20 is an expansion circuit for expanding compressed image data and restoring an image, and 22 is a feature of the present embodiment.
A recompression circuit 24 for recompressing the compressed image data by the compression circuit 18 is an image memory for storing the compressed image data. Reference numeral 26 denotes a modem for communicating with the outside via a public telephone line. The circuits or devices 10 to 26
Are interconnected via

FIG. 2 shows a compression circuit 1 according to the JPEG system.
8 and the circuit configuration of the expansion circuit 20. Compression circuit 18
Is a DCT circuit 30 that performs discrete cosine transform of a pixel block composed of 8 pixels × 8 pixels, a quantization circuit 32 that quantizes a transform coefficient output from the DCT circuit 30, and Huffman coding of an output of the quantization circuit 32. It comprises a Huffman encoder 34. The output of Huffman encoder 34 is applied to image memory 24 via bus 28 and stored.

The decompression circuit 20 includes a Huffman decoder 36 for Huffman decoding compressed image data, for example, compressed image data stored in the image memory 24, and an inverse quantization circuit 38 for inversely quantizing the output of the Huffman decoder 36. , And inverse DCT for inverse DCT transforming the output of the inverse quantization circuit 38
It comprises a circuit 40.

The CPU 10 sets the quantization and dequantization tables necessary for the quantization circuit 32 and the dequantization circuit 38 in the circuits 32 and 38, and stores the tables required for the Huffman encoder 34 and the Huffman decoder 36 in the circuits. 34,
Set to 36.

FIG. 3 shows data flows in DCT transform, quantization, Huffman coding, Huffman decoding, inverse quantization, and inverse DCT transform. The DCT circuit 30 has 8 rows and 8 rows.
DCT-transform the image data {sij} of the column, and transform coefficient {S
ij｝ is output. In {Sij}, i = 0 and j = 0 are DC coefficients, and the frequency increases toward the right side and the lower side.

The quantization circuit 32 quantizes {Sij} with a quantization table {Qij}, that is, divides Sij by Qij, and outputs a quantized DCT coefficient {Sqij}. Considering this, Qij is the lowest when i = 0 and j = 0, and increases as i and j increase, considering that the human visual characteristic has high sensitivity at low frequency and sensitivity decreases as the frequency increases. It is set to be.

The Huffman encoder 34 generates {Sqij}
Is encoded in Huffman encoding, and the encoded data is stored in the image memory 24 together with the quantization table {Qij}. The data thus stored in the image memory 24 is modulated by the modem 26 as necessary and transmitted to the transmission path.

The processing in the decompression circuit 20 is the reverse of the above. The Huffman decoder 36 Huffman-decodes the data received from the transmission line and outputs quantized DCT coefficients {Sqij}. The inverse quantization circuit 38 calculates the quantized DCT coefficient ｛S
qij} is inversely quantized using the same quantization table {Qij} as used for quantization, and DCT coefficients {Rij} are output. That is, Sqij is multiplied by Qij, and the multiplication result is set to Rij. The inverse DCT circuit 40 performs an inverse DCT transform on {Rij} and outputs restored image data {rij}.

As can be seen from FIG. 3, the compression ratio can be changed by the quantization table {Qij}. For example, when Qij is increased, SQij is decreased and the compression ratio is increased. However, the quantization error increases and the image quality generally deteriorates. That is, in the JPEG system, the image quality and the compression ratio have a trade-off relationship.

Referring to FIG. 4, Huffman encoder 3
4 will be described. The quantized DCT coefficients {Sqij} are divided into DC components (i = 0, j = 0) and other AC coefficients, and the AC coefficients are zigzag-scanned and rearranged in order from low-frequency components to high-frequency components.

As for the DC coefficient, the difference value Diff from the DC coefficient of the previous pixel block is grouped according to the magnitude of the value as shown in FIG. 5, the corresponding SSSS is Huffman-coded, and the actual value is calculated. It is represented by an additional bit.

On the other hand, regarding the AC coefficient, the run length of the zero coefficient (ineffective coefficient) up to the next non-zero coefficient ZZ (effective coefficient) is defined as RRRR, and as shown in FIG. Divide and RRRRSSSS is 2
Dimensional Huffman coding, and ZZ is represented by additional bits. Note that EOB and ZRL in the Huffman code table of AC coefficients are exceptional processing. EOB indicates the end of the block when all coefficients remaining in the block are 0, and ZRL is
Indicates that the run length of the zero coefficient is 16 or more.

FIG. 7 shows a circuit configuration of the recompression circuit 22.
FIG. 8 shows a change in the data, and FIG. 9 shows an operation flowchart.

First, the CPU 10 sets a Huffman table in the Huffman encoder 50 and the Huffman decoder 56, and sets a quantization table and a requantization table in the inverse quantization circuit 52 and the requantization circuit 54, respectively (S1). ). Thereafter, processing is sequentially performed in block units.

The Huffman decoder 50 is used for the image memory 2
Huffman decoding is performed on a part of the lock of the compressed image data stored in No. 4 and a quantized DCT coefficient {Sqij} is output (S2).

The inverse quantization circuit 52 converts the quantized DCT coefficient {Sqij} into the same quantization table {Qij} as used for quantization.
, And outputs a DCT coefficient {Rij}.
The requantization circuit 54 outputs the output {Rij} of the inverse quantization circuit 52.
Is again quantized by another quantization table {Qaij}. At this time, if it is desired to increase the compression ratio, Qaij ≧
Qij.

It is clear that the inverse quantization and requantization can be realized by one operation (S4). However, if the quantized DCT coefficient of interest is 0, it is also 0 by inverse quantization and requantization, so that inverse quantization and requantization are omitted (S3).

When the inverse quantization and requantization operations are performed on all 64 coefficients in one block (S5), the Huffman encoder 56 outputs the output ΔT of the requantization circuit 54.
qij｝ is Huffman-coded, and its output is stored in an image memory 24.
(S6).

The processing of S2 to S6 is executed for all blocks of one screen (S7).

As shown in the flowchart, 64
It is described that Huffman coding is performed after the operations of inverse quantization and requantization are completed for all the coefficients, but the operations of inverse quantization and requantization and Huffman coding are performed as simultaneously as possible. Of course.

It is clear that the operation Tqij = Sqij × Qij / Qaij shown in S4 can be realized only by multiplication or by shift operation. For example, if Qij and / or Qaij are selected so that Qij / Qaij is an appropriate positive value, only multiplication is performed, and if Qij / Qaij is a power of 2,
Only shift operation is required. This modification is shown in FIG. Between the Huffman decoder 50 and the Huffman encoder 56,
What is necessary is just to insert the coefficient operation circuit 58 which performs a multiplication or a shift operation. Obviously, the speed can be increased by using the multiplication or the shift operation.

Instead of the Huffman encoder 56 or in parallel, an encoder for another code may be arranged. Then, the recompression circuit 22 can also be used as a code converter.

FIG. 11, FIG. 12, and FIG. 13 are operation flowcharts for executing recompression of this embodiment.
12 shows a main routine, FIG. 12 shows a subroutine for processing the DC coefficient in FIG. 11, and FIG. 13 shows a subroutine for processing the AC coefficient.

[0034]

As described above, claim 1 of the present application is as described above.
According to the invention according to the invention, the image is orthogonally transformed, quantized, Huffman
Executes an encoding algorithm such as JPEG encoding.
Quantization without inverse orthogonal transform on devices that can perform
Decoding process of Huffman encoding accompanying the subsequent stage (at the time of encoding)
Inverse quantization and requantization are performed by
Or by performing a bit shift operation, etc.
Use a different quantization table than the one used for
Huffman coded compression results can be obtained at high speed.
You. In particular, according to the second aspect of the invention, in addition to the above effects,
The system has a separate compression unit and recompression unit.
Independent of the orthogonal transform, quantization, and Huffman coding in
Huffman decoding, different quantization tables in recompressor
Change of quantization output between Huffman coding
The non-execution time of the normal encoding process is
Regardless, it can be recompressed at the desired time.

[Brief description of the drawings]

FIG. 1 is a schematic block diagram of an embodiment of the present invention.

FIG. 2 is a detailed circuit block diagram of a compression circuit 18 and a decompression circuit 20.

FIG. 3 is an explanatory diagram of a change in data in compression / expansion.

4 is an explanatory diagram of a process in a Huffman encoder 34. FIG.

FIG. 5 is an explanatory diagram of encoding of a DC coefficient.

FIG. 6 is an explanatory diagram of encoding of an AC coefficient.

FIG. 7 is a detailed circuit block diagram of the recompression circuit 22.

FIG. 8 is an explanatory diagram of a data change in FIG. 7;

FIG. 9 is an operation flowchart of FIG. 7;

FIG. 10 is a modification example of FIG. 7;

FIG. 11 is a main routine of an operation flowchart for executing recompression according to the present embodiment.

FIG. 12 is a flowchart illustrating a subroutine for processing a DC coefficient in FIG. 11;

FIG. 13 shows a subroutine for processing an AC coefficient.

[Explanation of symbols]

10: CPU 12: Image Scanner 14: Printer 16: Image Display Device 18: Compression Circuit 20: Decompression Circuit 22: Recompression Circuit 24: Image Memory 26: Modem 28: Bus 30: DCT Circuit 32:
Quantization circuit 34: Huffman encoder 36: Huffman decoder 38: Inverse quantization circuit 40: Inverse DCT
Circuit 50: Huffman decoder 52: Inverse quantization circuit 54: Requantization circuit 56: Huffman encoder
58: Coefficient operation circuit

Claims (2)

(57) [Claims] (1)Conversion means for orthogonally transforming image data; A transform coefficient obtained by the orthogonal transform is converted to a first quantization table.
Quantization means for quantizing according to the A first output for Huffman encoding the output obtained by the quantization
Huffman encoding means, The compressed image data obtained by the Huffman coding is
Huffman decoding means for decoding The first quantization table obtained by the Huffman decoding
Is output according to the second quantization table.
Changing means for changing the output to a quantized output, Huffman coding of the output obtained by the quantization of the changing means
Second Huffman encoding means Characterized by
Recompression device. (2)A recompression system including an encoding unit and a recompression unit
System The encoding unit includes: Conversion means for orthogonally transforming image data; A transform coefficient obtained by the orthogonal transform is converted to a first quantization table.
First quantizing means for quantizing according to A first output for Huffman encoding the output obtained by the quantization
Huffman encoding means, The recompression unit, The compressed image data obtained by the Huffman coding is
Huffman decoding means for decoding The first quantization table obtained by the Huffman decoding
Is output according to the second quantization table.
Changing means for changing the output to a quantized output, Huffman coding of the output obtained by the quantization of the changing means
And a second Huffman encoding means.
Recompression system.