JP2003299083A - Encoder, data conversion apparatus, encoding program and data conversion program - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To perform rotation processing of an image in a short period of time by simplifying processing of a JPEG (joint photographic experts group) encoder for performing rotation processing of an image. <P>SOLUTION: When an input image data are subjected to rotation processing, after DCT (discreet cosine transform) conversion (B1) and quantization (B2), rotation zigzag processing (B6) is performed on the basis of information of a quantization table (B5) and information of a rotation angle of the image, and then the data are subjected to Huffman encoding (B4). In the rotation zigzag processing, a zigzag sequence is decided on the basis of the rotation information. Further, on the basis of the quantization table information, it is analyzed whether or not the quantization table is diagonally symmetric, and correction is performed on the basis of the result of analysis. <P>COPYRIGHT: (C)2004,JPO

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to a coding device, a data conversion device, a coding program and a data conversion program, and more particularly to a coding device, a data conversion device, and a coding device for rotating an image. Program and data conversion program.

[0002]

2. Description of the Related Art Conventionally, JPEG (Joint Photograph)
A method of compressing and coding an image by the ic Coding Experts Group) method is known.

In the JPEG system, a normal image is 8 ×
By dividing the block into 8 blocks of 64 pixels and performing DCT (Discrete Cosine Transform) transformation on the data of each block, a process for obtaining the same value (DCT coefficient) representing the magnitude of the frequency component as the number of pixels is performed. The DCT coefficient is quantized by the quantization table. The quantized DCT coefficient is subjected to a process of extracting coefficients in order from the lowest frequency by a zigzag sequence, and Huffman coding is performed. Information necessary for decoding, such as image size and color type, and table information used for quantization and encoding is added as header information (marker addition) to the beginning of the encoded signal to form a JPEG file. It

When decoding a JPEG file, a table is extracted from the file, decoding is performed by referring to the Huffman coding table, and inverse quantization is performed by referring to the quantization table. This gives the DCT coefficient,
Finally, the brightness of each pixel is obtained based on the DCT coefficient. As a result, the original image is restored.

In the case of a color image, the above process is performed for each component. The encoding process for generating the JPEG file as described above is as shown in FIG.

FIG. 11 is a flowchart showing the process of JPEG file generation. With reference to FIG. 11, a predetermined sub-sampling process is performed on the image data to be encoded. The block-and-level-shifted image data is subjected to DCT conversion processing by the DCT conversion unit (S1). After that, the DCT coefficient obtained by the DCT transform is quantized by a predetermined quantization table (S3). The quantized coefficients are rearranged in a predetermined order by the zigzag sequence section (S5), and then coded by the encoder to generate a bitstream (S7). A JPEG file is generated by adding a JPEG marker to this bitstream.

Further, for example, an MFP (Multi Function P
eripheral), copiers, printers, scanners, and other devices that handle image data may perform processing for changing the orientation of the image. When a JPEG file is generated by changing the orientation of an image in this way, the following processing is performed.

(1) From bitmap data to JPEG
When Generating a Rotation File FIG. 12 is a flowchart showing a process of generating a JPEG file in which an image is rotated from bitmap image data.

Referring to FIG. 12, the DCT domain processing is performed on the data (S101) after the DCT conversion.
The rotation process is performed for each block of × 8 (S10).
3). After that, quantization processing (S105), zigzag sequence processing (S107), Huffman coding (S109)
Then, the code data subjected to the rotation process in block units is stored in the memory (S111). By rearranging the stored code data (S113), code data in which the entire image is correctly rotated is generated.

A specific example of the rotation processing of the DCT block by the DCT domain processing will be described below. In the DCT domain processing, the rotation processing is performed in the 8 × 8 block by performing, for example, the following processing on the DCT coefficient obtained by the DCT conversion.

90 ° rotation (-1) ^v · F (u, v) 180 ° rotation (-1) ^u · (−1) ^v · F (u, v) 270 ° rotation (−1) ^u · F ( u, v) In the above processing, F (u, v) indicates a DCT coefficient (both u and v are 0 to 7, and each block forms a block of 8 × 8), and u is a DCT. A row of coefficients is shown, and v is a column of DCT coefficients.

That is, by the above processing, when the block is rotated by 90 °, the DCT coefficient of the odd column is multiplied by -1, and when it is rotated by 270 °, the DCT coefficient of the odd row is multiplied by -1. Is performed. Also,
In the case of 180 ° rotation, the DCT coefficient is -1 in a checkerboard pattern.
The process of multiplying is performed.

(2) J rotated from a JPEG file
When Generating a PEG File FIG. 13 is a flowchart showing a process when a rotated JPEG file is generated from a JPEG file.

As shown in FIG. 13, the input JP
The EG file is decoded by the Huffman decoder (S2
01). After the zigzag sequence of the decoded data is released (S203), inverse quantization is performed (S20).
5).

DCT domain processing is performed on the inversely quantized data (S207), whereby rotation processing is performed in 8 × 8 block units. After that, quantization processing (S209) and zigzag sequence processing (S21)
1), Huffman coding is performed (S213), and code data subjected to rotation processing in block units is stored in the memory (S215). By rearranging the stored code data, it is possible to generate code data in which the entire image is correctly rotated (S217).

[0016]

In the case of performing rotation processing on a JPEG file, in the prior art, the DCT domain processing is added to the DCT coefficient by going up to the inverse quantization processing. After that, it was necessary to perform quantization, coding and rearrangement processing on the data. Therefore, the conventional JP
In the device for rotating the EG file, the processing becomes complicated, and it takes a long time to obtain the desired rotated file.

Also, when a JPEG rotated file is generated from bitmap data, the processing becomes complicated, and it takes time to obtain the rotated file.

The present invention has been made to solve the above problems, and is simple in processing and capable of performing processing in a short time. Encoding apparatus, data conversion apparatus, encoding program and data conversion. The purpose is to provide the program.

[0019]

According to one aspect of the present invention to achieve the above object, an encoding device is a JPEG.
An encoding device for encoding, characterized in that a coefficient for rotation processing is determined based on quantization table information and rotation information of image data after quantization of image data and before encoding.

According to another aspect of the present invention, the data conversion apparatus adds rotation processing to the JPEG file and again executes JP
A data conversion device for outputting as an EG file,
It is analyzed whether or not the quantization table in the header of the JPEG file to be processed is diagonally symmetric, and if it is diagonally symmetric, inverse quantization and quantization processing of the JPEG file is not performed. And

According to still another aspect of the present invention, the data conversion device is a data conversion device that performs a rotation process on a JPEG file and outputs it as a JPEG file again, and the quantum data in the header of the JPEG file to be processed. It is characterized in that whether or not the quantization table is diagonally symmetric is analyzed, and if it is diagonally asymmetric, the inverse quantization process, the DCT domain process, and the quantization process are performed only on the asymmetric coefficient. And

According to another aspect of the present invention, the encoding program is an encoding program for performing JPEG encoding, and the quantization table information and the image data are rotated after the image data is quantized and before the encoding. Based on information and
It is characterized in that the coefficient of the rotation process is determined.

According to still another aspect of the present invention, the data conversion program is a data conversion program for rotating a JPEG file and outputting it again as a JPEG file, and the quantum data in the header of the JPEG file to be processed. It is analyzed whether or not the quantization table is diagonally symmetric, and if it is diagonally symmetric, the inverse quantization and the quantization processing of the JPEG file are not performed.

[0024]

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the drawings.

[First Embodiment] An encoding apparatus according to the first embodiment of the present invention rotates an image and executes JP
Output as an EG file. When performing the rotation process, the DCT domain process is performed in the zigzag sequence section based on the quantization table information and the rotation information (rotation angle). As a result, the rotation process can be performed by one process (here, the part that performs the zigzag sequence), and the rotation process can be speeded up. Further, when considering hardware implementation, it is possible to minimize an increase in circuit scale.

That is, in the coding apparatus according to the present embodiment, after the quantization of the image data and before the coding, based on the quantization table information and the rotation information of the image data,
The coefficient of the rotation process used in the zigzag sequence is determined, and the process is performed based on the coefficient.

FIG. 1 is a block diagram showing the arrangement of the encoding apparatus according to the first embodiment. Referring to the figure, the encoding device includes a DCT transform unit (B1) and a quantization unit (B2).
A zigzag processing unit (B3) that performs normal zigzag sequence processing when an image is not rotated, a Huffman coding unit (Huffman encoder) (B4), and a memory (B
7), a data rearrangement processing unit (B9), and a marker addition unit (B15).

A rotation zigzag processing unit for performing rotation processing and zigzag sequence processing when rotating an image based on the input unit (B5) for inputting quantization table information and the quantization table information and rotation information. (B
6) and 6) are provided in the encoding device.

The input image data to be processed is DC
DCT conversion is performed by the T conversion unit (B1). Then, the quantizing unit (B2) performs a quantizing process. When the image rotation is not performed, normal zigzag processing is performed by the zigzag processing unit (B3), encoding is performed by the Huffman encoding unit (B4), and a bitstream is generated. A marker is added to this bitstream by the marker addition unit (B15), and the bitstream is output as a JPEG file.

The processing for rotating an image will be described below. The image data is the DCT conversion unit (B1)
DCT conversion is performed by. Then, the quantizing unit (B2) performs a quantizing process. Up to this point, the process is the same as when the rotation process is not performed.

Here, information on the quantization table used for encoding and image rotation information (90 ° / 180 ° / 270).
The rotation zigzag processing is performed by the rotation zigzag processing unit (B6) on the basis of In this rotation zigzag processing, rotation processing in the DCT block is performed.
At the same time, zigzag sequence processing is performed. After that, the bitstream data encoded by the Huffman encoder (B4) is stored in the memory (B7), and the data rearrangement processing unit (B9) turns the entire image into encoded data. By adding a marker to this data, a rotated JPEG file can be obtained.

FIG. 2 is a diagram showing data output from the quantizer (B2). Referring to the figure, the quantizer (B
The output data from 2) is data composed of one block of 8 × 8, and the data at the upper left is D _0, and it is shown in the order from the right to the lower right by the variables D _{0 to} D ₆₃ .

FIG. 3 is a diagram showing a zigzag sequence which is performed in the zigzag processing section (B3) without rotation processing. As shown in the figure, from the lower frequency side, which is the upper left of the image, to the higher frequency side, which is the lower right, 0 →
In the order of 63, the DCT coefficient is replaced with the code. That is, the zigzag scan is performed from the low frequency side to the high frequency side.

The zigzag processing for rotation will be described in detail below. First, the zigzag sequence is determined from the image rotation information.

FIG. 4 is a diagram showing a zigzag sequence at 90 ° rotation, and FIG. 5 shows 270 ° rotation (left 90 °).
It is a figure which shows the zigzag sequence in the case of (rotation). With reference to the figure, these sequences are formed in that the sequences are formed from the upper left to the lower right of the DCT coefficient.
The sequence is the same as the sequence shown in Fig. 3 when no rotation is performed, but the zigzag swing is reversed between Figs. 3 and 4 and 5. In the case of 90 ° or 270 ° rotation, the quantized DCT coefficient is extracted in the sequence shown in FIG. 4 or 5.

In the case of 90 ° or 270 ° rotation, the zigzag swing is opposite to that in the case where no rotation is performed. Therefore, the quantization table information is referred to and the DCT coefficient is compared. The correction process to be performed is determined.

If the quantization table used is diagonally symmetric as shown in FIG. 6, the DCT coefficient corresponding to the hatched sequence in FIGS. 4 and 5 is multiplied by -1. Done. That is, as shown in FIG. 6, regarding an 8 × 8 block of DCT coefficients, an axis from the upper left to the lower right is assumed, and if the values in the quantization table are symmetric with respect to this axis, FIG. Only the coefficient corresponding to the hatched sequence of No. 5 is multiplied by -1, and no particular correction of the quantization coefficient is performed.

More specifically, diagonal symmetry means that each coefficient of quantization is symmetric with respect to the broken line (diagonal line) of FIG. 6, and Q ₁ = Q ₈ , Q ₂ = Q ₁₆ , Q ₁₀ = Q ₁₇ , Q ₃ = Q ₂₄ ,
.., Q ₅₅ = Q ₆₂ .

On the other hand, as shown in FIG. 7, the coefficient of the quantization table is asymmetric with respect to the axis from the upper left to the lower right, that is, when the value of Q ₃ is different from the value of Q ₂₄ , or Q ₅ If the value of is different from the value of Q _40, the coefficient corresponding to this asymmetric portion is corrected.

Here, the correction processing applied to each coefficient from the quantization table information will be described. It is assumed that the value of Q ₃ shown by hatching in FIG. 7 is different from the value of Q _{24 and} the value of Q ₅ is different from the value of Q ₄₀ . Here, assuming that the image is rotated by 90 °, the data processed in the asymmetric part of the DCT coefficient is D ₃ , D ₂₄ , D ₅ and D _{40 in} FIG. 2 (quantized DCT coefficient). Becomes When performing zigzag processing for rotation on these data, each (Q
_{3 / Q 24), (Q} 24 / Q 3), (Q 5 / Q 40) and (Q
Multiply by the correction factor of ₄₀ / Q ₅ ). As a result, it is possible to correct the DCT coefficient and prevent a mismatch with the quantization table at the time of decoding.

FIG. 8 is a diagram showing a zigzag sequence when performing 180 ° rotation. When rotating by 180 °, the zigzag sequence is the same as the sequence when not rotating (FIG. 3). However, since the image is rotated by 180 °, a process of multiplying the DCT coefficient of the hatched portion in FIG. 8 by -1 in the sequence is performed. When the image rotation is performed as described above, the data processed by the rotation zigzag processing unit (B6) is encoded by the Huffman encoding unit (B4) and temporarily stored in the memory (B7).

Since the coded data in the memory is only rotated in block units, rearranging them makes it possible to obtain coded data in which the entire image is rotated. A marker such as a header or table information is added to the rearranged code data in the marker adding unit (B15), thereby creating a rotated JPEG file.

[Second Embodiment] FIG. 9 is a block diagram showing the structure of a data conversion apparatus according to the second embodiment of the present invention. Referring to the figure, the data conversion device is
Huffman decoder for inputting PEG files (B10)
A rotation zigzag processing unit (B11), a Huffman encoding unit (Huffman encoder) (B12), and a memory (B1
3), a data rearrangement processing unit (B14), a marker addition unit (B15), and a quantization table analysis unit (B16).

The data conversion apparatus according to the second embodiment uses a JPEG file as an input and rotates the JP.
Output EG file. The input JPEG file is decoded by the Huffman decoder (B10). The decrypted data is passed to the rotation zigzag processing unit (B11).

Here, the quantization table analysis unit (B16)
At, the quantization table in the file is analyzed, and the re-encoding order is determined based on the information and the rotation information. That is, since the data decoded by the Huffman decoder (B10) are arranged in a normal zigzag sequence, when 90 ° rotation or 270 ° rotation is performed, the DCT coefficients are set in the order shown in FIGS. 4 and 5. The sorting process is performed.

Further, here, the quantization table information of the JPEG file is analyzed by the quantization table analysis unit (B16), and the processing to be applied to each DCT coefficient is determined from the analysis result. The quantization table analysis unit (B16) analyzes the following items.

Check whether the quantization table used in the JPEG file is diagonally symmetric as shown in FIG.

If it is not diagonally symmetric (in the case of the configuration shown in FIG. 7), asymmetric position information is acquired.

As a result of the analysis, if the quantization table to be used is diagonally symmetric as shown in FIG. 6, the DCT coefficient corresponding to the sequence of the hatched portions in FIGS. 4 and 5 is multiplied by -1, and the Huffman coding unit The data is passed to (B12) and re-encoding is performed. If it is not diagonally symmetric,
Correction is performed on the DCT coefficient that hits the asymmetric portion.

Also in this correction, as in the first embodiment, the following correction is performed on the DCT coefficient. For example, assume that the value of Q _{3 and} the value of Q ₂₄ are different, and the value of Q _{5 and} the value of Q ₄₀ are different, as shown in FIG. 7. Here, if the output data of the Huffman decoder (B10) is the one shown in FIG. 2, the data of the asymmetric portion becomes D ₃ , D ₂₄ , D ₅ and D ₄₀ when 90 ° rotation is performed. When performing zigzag processing for rotation on these data, (Q ₃ / Q ₂₄ ), (Q ₂₄
/ Q ₃ ), (Q ₅ / Q ₄₀ ) and (Q ₄₀ / Q ₅ ) correction factors.

Similarly, when 180 ° rotation is performed, the decoded DCT coefficient (quantized DCT coefficient) is taken out in accordance with the zigzag sequence shown in FIG. The coefficient corresponding to the hatched sequence is multiplied by -1.

The data subjected to the zigzag processing for rotation is re-encoded by the Huffman encoder (B12) and stored in the memory (B12).
It is accumulated in 13). Since the code data in the memory is only rotated in block units, the data rearrangement unit (B14) rearranges the code data to generate code data in which the entire image is rotated. A rotated JPEG file can be obtained by adding a marker to the rearranged code data at the marker addition unit (B15).

When rotating the JPEG file and outputting the JPEG file again, it is analyzed whether the quantization table in the header of the JPEG file to be processed is diagonally symmetric and the diagonal In the case of symmetry, dequantization and quantization processing of the JPEG file is not performed, rotation is performed by zigzag processing for rotation, and if the quantization table is not diagonally symmetric, rotation processing is performed by processing equivalent to the conventional technology. You may do it.

That is, after the decoding by the Huffman decoder (B10) as shown in FIG. 10, if the quantization table is diagonally symmetric, the processing shown by B11 to B15 in FIG. If not, S203 to S in FIG.
Processing is performed in blocks B21 to B25 corresponding to 211, and thereafter, processing is performed in blocks B12 to B14 corresponding to S213 to S217 in FIG.

Similarly, when the JPEG file is rotated and output again as a JPEG file, it is analyzed whether the quantization table in the header of the JPEG file to be processed is diagonally symmetric, If it is diagonally asymmetric, the inverse quantization process, the DCT domain process, and the quantization process may be performed only on the asymmetric coefficient.

As described above, according to the embodiment of the present invention, in the generation of the JPEG file, the D-pattern is zigzag based on the quantization table information and the rotation information.
Rotation processing based on CT domain processing can be performed. As a result, the JPEG image rotated from the input image data
The file creating system can perform the same processing in the data processing from DCT transformation to quantization and the Huffman coding processing regardless of whether the image is rotated or not.

Also, in a system in which a JPEG file is rotated and output again as a JPEG file, the inverse quantization / quantization processing can be minimized based on the analysis result of the quantization table. Alternatively, the inverse quantization / quantization process can be omitted. This makes it possible to provide an encoding device that is simple in processing and can be processed in a short time.

The processing in the above embodiment is
This may be done by software or by using a hardware circuit.

It is also possible to provide a program for executing the processing in the above-mentioned embodiment, and to record the program in a recording medium such as a CD-ROM, a flexible disk, a hard disk, a ROM, a RAM or a memory card. It may be provided to the user. Further, the program may be downloaded via a communication line such as the Internet.

The present invention can be applied to a computer system connected to a network and a computer system not connected to a network environment.

The embodiments disclosed this time are to be considered as illustrative in all points and not restrictive. The scope of the present invention is shown not by the above description but by the claims, and is intended to include meanings equivalent to the claims and all modifications within the scope.

[Brief description of drawings]

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

FIG. 2 is a diagram showing output data from a quantization unit.

FIG. 3 is a diagram showing a zigzag sequence without rotation processing.

FIG. 4 is a diagram showing a zigzag sequence in 90 ° rotation.

FIG. 5 is a diagram showing a zigzag sequence at 270 ° rotation.

FIG. 6 is a diagram for explaining a case where a quantization table has diagonal symmetry.

FIG. 7 is a diagram for explaining a case where a quantization table is diagonally asymmetric.

FIG. 8 is a diagram showing a zigzag sequence in the case of 180 ° rotation.

FIG. 9 is a block diagram showing a configuration of a data conversion device according to a second embodiment.

FIG. 10 is a diagram showing a modification of the apparatus of FIG.

FIG. 11 is a flowchart showing a process of generating a conventional JPEG file.

FIG. 12 is a flowchart showing a process of generating a JPEG rotation file from a bitmap.

FIG. 13 is a flowchart showing a process of generating a rotated JPEG file from a JPEG file.

[Explanation of symbols]

B1 DCT transform unit, B2 quantization unit, B3 zigzag processing unit, B4 Huffman coding unit, B5 quantization table information input unit, B6 rotation zigzag processing unit, B7 memory, B9 data rearrangement processing unit, B10 Huffman decoder , B11 zigzag processing unit for rotation, B12 Huffman encoder, B13 memory, B14 data rearrangement processing unit, B15 marker addition unit, B16 quantization table analysis unit.

   ─────────────────────────────────────────────────── ─── Continued front page    F term (reference) 5B057 CD04 CG01 CH01 CH07 CH11                 5C059 MA00 MA23 MC02 MC14 RB09                       UA02 UA39                 5C076 AA24 BA06 BA09                 5C078 AA04 BA57 CA31 DA01 DB01

Claims (5)

[Claims] 1. An encoding device for performing JPEG encoding, wherein a coefficient of rotation processing is determined based on quantization table information and rotation information of image data after quantization of image data and before encoding. An encoding device characterized by the above. 2. A data conversion device which outputs a JPEG file again by applying rotation processing to the JPEG file, and analyzes whether or not the quantization table in the header of the JPEG file to be processed is diagonally symmetric. If it is diagonally symmetric, the data conversion device is characterized in that the inverse quantization and the quantization processing of the JPEG file are not performed. 3. A data conversion device that outputs a JPEG file again by applying rotation processing to the JPEG file, and analyzes whether the quantization table in the header of the JPEG file to be processed is diagonally symmetric. If it is diagonally asymmetric, the inverse quantization process is performed only on the asymmetric coefficient,
A data conversion device characterized by performing DCT domain processing and quantization processing. 4. A coding program for performing JPEG coding, wherein a coefficient of rotation processing is determined based on quantization table information and rotation information of image data after quantization of image data and before encoding. An encoding program characterized by the above. 5. A data conversion program for rotating a JPEG file and outputting it again as a JPEG file, and analyzing whether or not the quantization table in the header of the JPEG file to be processed is diagonally symmetric. If it is diagonally symmetric, the data conversion program is characterized in that the inverse quantization and the quantization processing of the JPEG file are not performed.