JP2000049621A - Huffman decoding method and data processor - Google Patents

Abstract

PROBLEM TO BE SOLVED: To speed up a decoding processing. SOLUTION: A decoding table forming means 171 generating the first table of a plurality of bit lengths, which corresponds to a Huffman code value by one to one, and the second table of cut width larger than the first table by using a plurality of bits for the respective inversion positions of the bits of the Huffman code value and a decoding means 173 inverting inputted data, obtaining the bit inversion positions and executing decoding by using the first or the second table in accordance with the bit inversion position are installed. The bit inversion position of inputted data is obtained and the second table is used in accordance with the bit inversion position. Thus a decoding processing speed is improved.

Description

DETAILED DESCRIPTION OF THE INVENTION

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image processing technique and, more particularly, to an improvement of a circuit for performing Huffman decoding.
For example, the present invention relates to a technology effective when applied to a data processing system of a digital still camera.

[0002]

2. Description of the Related Art For example, DCT (Discrete C) is an international standard for information encoding (information compression) technology for efficiently realizing communication of multimedia information such as image information and audio information.
JPE by osineTransfer (discrete cosine transform) method
G (Joint Photographic Expert Group). The encoder performs discrete cosine transform (DCT) on the information of the original image,
After this is quantized, for example, Huffman coding is performed so that the data encoded together with the parameters can be output to the transmission path. When the data coded together with the parameters is supplied from the transmission path, the decoder performs Huffman decoding on the data, performs inverse quantization, and then performs inverse discrete cosine transform (ID
CT) to generate a reproduced image. In quantization and dequantization, and in Huffman encoding and Huffman decoding, predetermined tables are respectively referred to.

As an example of a document describing Huffman encoding and Huffman decoding, in December 1991,
There is an “interface (from page 160)” issued by CQ Publishing.

[0004]

In the conventional Huffman decoding process, data that has been Huffman-encoded is compared in units of one bit. That is, it is determined whether the logical value of the first bit of the input data is “1” or “0”. If the logical value is “0”, the logical value of the next 1 bit is compared. In this way, when a different logical value appears, decoding is performed by returning to the previous bit. If the input data is 16 bits long,
Such a process is repeated 16 times, and it takes a very long time as a whole. Further, if a 16-bit table having a one-to-one correspondence with the Huffman code value can be used, the speed can be increased. However, forming such a table significantly increases the table capacity. -Considering application to a camera or the like, it is difficult to realize such a 16-bit length table within a limited memory capacity.

An object of the present invention is to provide a technique for speeding up Huffman decoding processing.

Another object of the present invention is to provide a technique for speeding up Huffman decoding without significantly increasing the table capacity.

[0007]

The following is a brief description of an outline of a typical invention among the inventions disclosed in the present application.

That is, a first table having a plurality of bits having a one-to-one correspondence with the Huffman code value is created, and a plurality of bits are used for each bit inversion position of the Huffman code value.
A plurality of second tables are created with a larger step width than the table, a bit inversion position of input data is obtained, and the data is converted using the first table or the second table according to the bit inversion position. Decrypt.

According to the above-mentioned means, the bit inversion position of the input data is obtained, and the second table is used according to the bit inversion position. Both the first table and the second table are tables having a plurality of bits, and the decoding process can be performed in units of a plurality of bits. This achieves a high-speed decoding process.

[0010] A fourth step (step S13) for determining whether or not the input data is larger than a predetermined value is provided.
It is preferable that the first table and the second table are selectively used according to the determination result of the fourth step.

When the input data is larger than a predetermined value, a fifth step (step S15) for obtaining a bit inversion position from the head of the data and a sixth step for shifting the data according to the bit inversion position (Step S15) and a seventh step (Step S16) of adding an offset value according to the bit inversion position can be provided.

Further, when the data processing device is configured to include a Huffman decoding unit for decoding the Huffman-encoded data, the Huffman decoding unit includes a plurality of bits corresponding to the Huffman code value on a one-to-one basis. Decoding table forming means (171) for creating a plurality of second tables with a step width larger than the step width of the first table using a first table having a long length and a plurality of bits for each bit inversion position of the Huffman code value. ) And decoding means (173) for inverting the input data to obtain a bit inversion position and decoding using the first table or the second table in accordance with the bit inversion position. .

In the above data processing apparatus, a discrete cosine transform unit (13) for performing discrete cosine transform of image data, a quantization unit (14) for quantizing a transform result of the discrete cosine transform unit, A Huffman encoding unit (1) for Huffman encoding the output data of the quantization unit.
5), an inverse quantization unit (18) for inversely quantizing the output data of the Huffman decoding unit, and an inverse discrete cosine transform unit (1) for performing an inverse discrete cosine transform of the output data of the inverse quantization unit.
9) can be provided.

[0014]

FIG. 1 shows a configuration example of a digital still camera which is an example of an image processing apparatus according to the present invention.

[0015] CCD (Charge Coupled)
Image data captured by the device camera 11 is converted from the RGB format to the YUV format, and then stored in the image memory 12 at the subsequent stage. The YUV data read from the image memory 12 is input to the DCT unit 13 at the subsequent stage and is subjected to discrete cosine transform. The result of this conversion is quantized by the subsequent quantization unit 14, and then input to the subsequent Huffman encoding unit 15 and subjected to Huffman encoding. The Huffman encoded data is a JPEG file.

The data stored in the flash memory 16 is read out as needed, and is read out by the Huffman decoding unit 1.
7 where it is decoded. This decoding processing will be described later in detail. The data decoded by the decoding unit 17 is inversely quantized by the inverse quantization unit 18 and
After the inverse discrete cosine transform at T19, the subsequent liquid crystal display (LCD) 21 via the subsequent image memory 20
Is transmitted to and displayed.

The JPEG file contains a DHT marker and a DQT marker. The DHT marker is decoding table information, and the Huffman decoding unit 17 forms a decoding table based on the DHT marker, and performs Huffman decoding processing by referring to the table. The DQT marker is quantization table information, and the inverse quantization unit 18 creates an inverse quantization table based on the DQT marker, and performs an inverse quantization process by referring to the table.

FIG. 2 shows a configuration example of the Huffman decoding unit 17.

As shown in FIG. 2, the Huffman decoding unit 17 has a one-to-one correspondence with the Huffman code value, although there is no particular limitation.
And a decoding table forming means for generating a plurality of second tables having a larger step width than the first table by using a first table having a length of 16 bits and a plurality of bits for each bit inversion position of the Huffman code value. 171 and the first
A RAM (random access memory) 172 in which the table and the second table are stored, and a bit inversion position obtained after inverting input data (Huffman code value). Decoding means 173 for decoding using the table or the second table
And

As described above, the Huffman decoding process in the Huffman encoder 17 includes a step of creating a decoding table and a step of decoding with reference to the table.

Next, the decoding table creation performed by the decoding table creation means 171 will be described.

First, the inverted values of the Huffman code value (16-bit length) are arranged in ascending order. If the Huffman code value is less than 16 bits, it is left-justified data, and other than the code, the logical value "0" is assigned to make it 16 bits long.

As shown in FIG. 3, H'0 in hexadecimal notation is used for inverted values (indicated by 1) arranged in ascending order.
In the range of 01-H'007F (bit inversion positions: 0 to 6), a one-to-one table (indicated by 2) in increments of +1 is created, and H'0080-H'FFFF in hexadecimal notation.
In the range of (bit inversion position: 7-15), a plurality of tables (indicated by 4) are created using the next 4 bits (indicated by 3) in bit inversion position units. How many tables are needed is determined by the DHT marker of the Huffman coded data.

Next, a decoding process performed with reference to the decoding table created as described above will be described. This decoding processing is performed by the decoding processing means 172.

As shown in FIG. 3, a 16-bit Huffman code value is read from the flash memory 16 and
Bit invert the data value. Then, a bit inversion position (indicated by 5) is obtained from the bit-inverted data.

When the bit inversion position is the 0th to 7th bits, the inverted value (indicated by 6) is used as a key (indicated by FIG. 7), and a search of the decoding table is performed. Table values are determined.

If the bit determination position is the 8th to 15th bit, as shown in FIG.
The bit (indicated by 8) is taken out, and an offset value (indicated by 9) corresponding to the bit inversion position is added to obtain a corresponding table value as a key for decoding (indicated by 10).

FIGS. 6 and 7 show more specific decoding tables.

Due to space limitations, FIG. 1 to N
o. 223 is shown in FIG. 224
-No. 319 is shown. Further, portions where the values are continuous are partially omitted.

FIGS. 6 and 7 show table values, their corresponding signs, and their inverted values. For convenience of explanation, the sign and its inverted value are shown in both binary and hexadecimal. The table value contains the group number (step SSS
S), code length (length), and run length (run).

Inversion values of the Huffman code value (16-bit length) are arranged in ascending order, and H'000-H'00 in hexadecimal notation.
In the range of 7F (bit inversion positions: 0 to 6), a one-to-one table (first table) in increments of +1 is set, and H'0080-H'FFFF in hexadecimal notation (bit inversion positions: 7).
In the range of -15), a plurality of tables (second tables) are created with a larger step width than the first table using the next 6 bits for each bit inversion position. The step width in the second table is as follows.

No. In the case of Nos. 128 to 160, the increment is H'02, In Nos. 161 to 191, the unit is H′02,
In Nos. 192 to 207, H'10 increments are used. 208-2
No. 23, H'20 increments; In 224 to 239,
No. H'40, No. In the case of Nos. 240 to 255, H'80 is used. In 256-271, H'100 increments, N
o. In Nos. 272 to 287, H'200 increments are used. 28
In Nos. 8 to 303, steps of H'400 are used. 304-31
In H.9, the increment is H'800.

FIG. 8 shows a specific flow of the decoding process by the decoding means 173.

First, a 16-bit Huffman code value is read from the flash memory 16 (step S11), and the data value is bit-inverted (step S12).

Next, it is determined whether the data bit-inverted in step S12 is greater than H'80 in hexadecimal notation (step S13). If it is determined in this determination that the value is smaller than H'80 (NO), the decoding table is searched using the bit-inverted data as a key, and the corresponding table value (S
SSS, length, run) are obtained.

In the determination in step S13, the inverted data is larger than H'80 (YES).
If it is determined that the bit inversion position is obtained from the beginning (step S14), the bit is shifted to the right according to the bit inversion position (step S15), and an offset value is added according to the bit inversion position. Is used to obtain a key (step S16). Here, the second table (N
o. The keys in 128 to 319) are as follows.

No. In 128 to 160, the inverted value is shifted right by one bit, and the offset value H'40 is added, so that the number of keys becomes 0080 to 00A0.

No. In 161 to 191, similarly, the inverted value is right-shifted by one bit and the offset value H′40 is added, so that the key becomes 3 from 00A1 to 00BF.
It is one.

No. In 192 to 207, the inverted value is right-shifted by 4 bits, and the offset value H'B0 is added, so that 16 keys from 00C0 to 00CF are provided.

No. In 208 to 223, the inverted value is right-shifted by 5 bits and the offset value H'C0 is added, so that 16 keys from 00D0 to 00DF are provided.

No. In 224 to 239, the inverted value is right-shifted by 6 bits, and the offset value H'D0 is added, so that 16 keys from 00E0 to 00EF are provided.

No. In 240 to 255, the inverted value is shifted right by 7 bits, and the offset value H'E0 is added, so that the keys are set to 16 keys from 00F0 to 00FF.

No. In 256 to 271, the inverted value is shifted right by 8 bits, and the offset value H′F0 is added, so that 16 keys from 0100 to 010F are provided.

No. In 272 to 287, the inverted value is shifted right by 9 bits, and the offset value H'100 is added, so that the keys are 16 from 0110 to 011F.

No. In the case of 288 to 303, the inversion value is 10
The key is shifted right by a bit and the offset value H′110 is added, so that the key becomes 16 from 0120 to 012F.
Individual.

No. In 304 to 319, the inverted value is 11
The key is shifted right by a bit and the offset value H′120 is added, so that the key becomes 16 from 0130 to 013F.
Individual.

When the table is searched using such a key, the table value (SSSS, length, r
un).

Here, according to the prior art, the Huffman-encoded data is decoded by comparing the data in 1-bit units, so that when the input data is 16 bits long, The process is repeated 16 times, which is a very time-consuming process. On the other hand, in the above example, the Huffman encoded data is
Since the decoding process is performed on a bit-by-bit basis, data comparison is performed on a bit-by-bit basis as in the related art.
Decoding efficiency can be improved.

The use of a 16-bit table having a one-to-one correspondence with the Huffman code value makes it possible to increase the speed. On the other hand, when such a table is formed, the table capacity is significantly increased. Digital still
Considering application to a camera or the like, it is difficult to realize such a 16-bit length table within a limited memory capacity. On the other hand, in the above example, when the Huffman-encoded data is less than H'80, the first table is searched using the data (inversion value) as a key. Data is H '
When the number becomes 80 or more, the second table having a step size larger than the first table is searched, and all of the data is one-to-one with Huffman-encoded data.
Therefore, the table capacity does not need to be significantly increased because the table is not 16 bits long. This is advantageous in a device such as a digital still camera, which cannot be equipped with a large-capacity RAM because of the manufacturing cost and space.

According to the above example, the following functions and effects can be obtained.

(1) First table (No. 1 to 127)
And the second table (No. 128 to 319) are both 1
Since the table has a 6-bit length and the decoding process can be performed in 16-bit units, the speed of the decoding process can be increased as compared with the case where the decoding process is performed in 1-bit units.

(2) H'000-H'00 in hexadecimal notation
In the range of 7F (bit inversion positions: 0 to 6), a one-to-one table in increments of +1 is used, but the Huffman code value H'0
In 080-H'FFFF, a plurality of second tables are created with a larger step width than the first table by using a plurality of bits for each bit inversion position. The storage capacity required for forming the table can be reduced as compared with the case of FIG.

Although the invention made by the inventor has been specifically described above, the present invention is not limited to this, and it goes without saying that various modifications can be made without departing from the gist of the invention.

For example, in the above example, the inverted value of the Huffman code value is obtained, and the key of the table search is obtained based on this inverted value. However, the key of the table search can be obtained without inverting the Huffman code value. You can ask.
Also, the boundary between the first table and the second table is H'80.
However, this value can be set arbitrarily. Further, the step width in the second table is not limited to the case shown in FIGS. 6 and 7, but may be changed as appropriate.

In the above description, mainly the case where the invention made by the present inventor is applied to a digital still camera, which is the field of application as the background, has been described.
The present invention is not limited to this, and can be applied to various data processing devices.

The present invention can be applied on condition that at least Huffman coded data is handled.

[0057]

The effects obtained by typical ones of the inventions disclosed in the present application will be briefly described as follows.

That is, the first table and the second table are both tables having a plurality of bits, and the decoding process can be performed in units of a plurality of bits. The speed of the decoding process can be increased.

In the second table, the step width of the table is set to be larger than +1 and a plurality of tables are formed. The storage capacity required for a large-capacity RAM, such as a digital still camera, can be reduced in terms of manufacturing cost and space.
This is advantageous in a device that cannot be equipped with a.

[Brief description of the drawings]

FIG. 1 is a block diagram illustrating a configuration example of a digital still camera to which a Huffman decoding method according to the present invention is applied.

FIG. 2 is a block diagram illustrating a configuration example of a Huffman decoding unit included in the digital still camera.

FIG. 3 is an explanatory diagram of table creation in the Huffman decoding method.

FIG. 4 is an explanatory diagram of a procedure for obtaining a table search key in the Huffman decoding method.

FIG. 5 is an explanatory diagram of a procedure for obtaining a table search key in the Huffman decoding method.

FIG. 6 is an explanatory diagram of a specific table in the Huffman decoding method.

FIG. 7 is an explanatory diagram of a specific table in the Huffman decoding method.

FIG. 8 is a flowchart showing a processing flow of the Huffman decoding method.

[Explanation of symbols]

 Reference Signs List 11 CCD camera 12, 20 Image memory 13 DCT unit 14 Quantization unit 15 Huffman encoding unit 16 Flash memory 17 Huffman decoding unit 18 Inverse quantization unit 19 IDCT unit 21 LCD 171 Decoding table forming unit 172 RAM 173 Decoding unit

 ──────────────────────────────────────────────────続 き Continued on the front page F term (reference) 5J064 AA03 BA09 BA16 BC01 BC02 BC16 BC23 BD03

Claims (5)

[Claims] 1. A Huffman decoding method for decoding Huffman-encoded data, wherein a first bit having a plurality of bits corresponding to a Huffman code value on a one-to-one basis.
A first step of creating a table, and a second step of creating a plurality of second tables with a step width larger than the step width of the first table using a plurality of bits for each bit inversion position of the Huffman code value; A third step of determining a bit inversion position of the input data and decoding the data using the first table or the second table according to the bit inversion position. Method. 2. The method according to claim 1, further comprising: a fourth step of determining whether or not the input data is greater than a predetermined value, wherein the first table and the second table are selectively used according to a result of the determination in the fourth step. 2. The Huffman decoding method according to 1. 3. When the input data is larger than a predetermined value, a fifth step of obtaining a bit inversion position from the head of the data, a sixth step of shifting data according to the bit inversion position, The Huffman decoding method according to claim 2, further comprising: a seventh step of adding an offset value according to the position. 4. A data processing apparatus including a Huffman decoding unit for decoding Huffman-encoded data, wherein the Huffman decoding unit has a plurality of bit length first tables corresponding to Huffman code values on a one-to-one basis. Decoding table forming means for generating a plurality of second tables with a step width larger than the step width of the first table using a plurality of bits for each bit inversion position of the Huffman code value; Decoding means for inverting to obtain a bit inversion position and decoding using the first table or the second table in accordance with the bit inversion position. 5. A discrete cosine transform unit for performing discrete cosine transform of image data; a quantizing unit for quantizing a transform result of the discrete cosine transform unit; and a Huffman encoding of output data of the quantizing unit. A Huffman encoding unit, an inverse quantization unit that inversely quantizes the output data of the Huffman decoding unit, and an inverse discrete cosine transform unit that performs an inverse discrete cosine transform of the output data of the inverse quantization unit. Item 5. The data processing device according to Item 4.