DE112008003626T5 - Low complexity coding in data compression systems - Google Patents

Abstract

Apparatus comprising:
an integrated circuit (IC) configured to compress and / or decompress data, the integrated circuit comprising an entropy coder configured to receive at least one discrete cosine transform (DCT) coefficient, wherein the entropy coder has a first Huffman table corresponding to a number of bits used to encode an amplitude of a non-zero AC coefficient.

Description

area

The The present disclosure describes a method of coding low complexity for use in data compression.

background

data compression Can be used to make a lot of information so space efficient as possible to represent. A data compression code can a mapping between a set of source messages and a Provide set of codewords. Some video compression systems can do this coding for a limited time Time span. For example, for high-definition television (HDTV) 1080i (1080 × 1920 30 fps) passing through tiles (ie rectangular fields) 8 × 1920 size, each encoded Tile be coded in 497 microseconds. Once encoded a tile is a subsequent tile at the input of the compression system be ready and ready for processing.

in the In general, entropy coding refers to a lossless one Data compression scheme, that of the special characteristics of the medium can be independent. Typically, entropy coding one of the most computationally complex elements a video compression system. In some cases Entropy coding involves the assignment of codes to symbols, code lengths with the probabilities of the symbols match. The design of a data compression system closes typically a compromise between effective compression of input data and minimum complexity costs in the Design.

entropy can be used in modern video compression systems using a Variety of techniques are realized, such as pairwise run lengths (run length (RLE)) and variable length coding (variable length coding (VLC)) or arithmetic coding. Arithmetic coding can provide a better compression rate, but can be very complex can not be a good solution for systems lower Be complexity. VLC-based techniques can provide a smaller compression rate and tend to do a lot less complex, therefore, these techniques can be used for Real-time applications may be more appropriate.

Brief description of the drawings

characteristics and advantages of the claimed subject matter are as follows detailed description of embodiments that come with it are consistent, with the description by reference should be considered in the attached drawings, in which:

1 Fig. 12 is a diagram showing an example of discrete cosine transform (DCT) based encoder processing;

2 Fig. 12 is a diagram showing an example of DCT-based decoder processing;

3 FIG. 12 is a diagram of an exemplary embodiment of low complexity entropy coding according to the present disclosure; FIG.

4 Figure 12 is a graph showing a rate-distortion curve (ie, compression rate versus PSNR) associated with standard JPEG entropy coding, as may be compared to entropy coding described in accordance with the present disclosure;

5 Figure 12 is a graph showing an estimated probability distribution of RUNLENGTH values in a standard JPEG algorithm;

6 is a graph showing an estimated probability distribution of SIZE values in a standard JPEG algorithm;

7 Fig. 10 is a flowchart of operations showing standard entropy coding using the JPEG image compression algorithm;

8th FIG. 10 is a flowchart of operations showing entropy coding using short tables of the RUNLENGTH according to an exemplary embodiment of the present disclosure; FIG.

9 FIG. 10 is a flowchart of operations according to another exemplary embodiment of the present disclosure; FIG. and

10 FIG. 10 is a flowchart of operations in accordance with an embodiment of the present disclosure. FIG.

Also if the following detailed description with reference continuing to illustrative embodiments, Professionals will find many alternatives, modifications and variations be the same apparent.

Detailed description

In general, this disclosure describes a method of encoding with low complexity for use in data compression. The embodiments described herein may be in accordance with data compression standards, such as Joint Photographic Experts Group (JPEG) standard, ITU-T T.81 . ISO / IEC IS 10918-1 , published in September 1992, can be used. More specifically, embodiments described herein can provide a number of entropy coding algorithms that can be used to reduce the memory space required for Huffman tables using a codec program.

data transfer of still images in many fields, such as color facsimile, desktop publishing and medical imaging can be done according to the JPEG standard. The JPEG standard can the use of discrete cosine transform (DCT) coding during include data compression processing.

Discrete Cosine Transform (DCT) based compression may include compressing a stream of 8x8 blocks of grayscale sample images. Each 8x8 block of source image samples is effectively a 64-point discrete signal, which may be a function of the two spatial dimensions x and y. The DCT coefficient values may therefore be considered as the relative amount of 2D spatial frequencies contained in the 64-point input signal. The zero frequency coefficient in both dimensions is referred to as the "DC coefficient", and the remaining 63 coefficients are referred to as the "AC coefficients". Additional information about the DCT is discussed in more detail "The JPEG Still Picture Compression Standard", Gregory K. Wallace, IEEE Transactions on Consumer Electronics, 1991 , described.

With reference to 1 now becomes a DCT-based encoder 100 shown. DCT-based encoder 100 can be configured to store some source image data (such as in 1 8 × 8 blocks shown) and output this data in a compressed form. encoder 100 can forward DCT (FCDT) 102 , Quantizer 104 and entropy coders 106 exhibit. In contrast, shows 2 a DCT-based decoder 200 which receives the compressed image data and outputs the original source image data in a reconstructed form. decoder 200 can be considered a reflection of encoder 100 As such, decoders can be considered 200 entropy 202 , Dequantizer 204 and inverse DCT (IDCT) 206 exhibit.

Again with reference to 1 can coder 100 be configured to source image data 101 for subsequent compression. Source image data 101 may be grouped in 8x8 blocks that are shifted from unsigned integers and in forward DCT 102 be entered. As discussed above, the specific operation of the forwarding DCT 102 in more detail in "The JPEG Still Picture Compression Standard", Gregory K. Wallace, IEEE Transactions to Consumer Eletronics, 1991 , described. FDCT 102 can be an output (for example, 64 DCT coefficients) for a quantizer 104 provide. In some embodiments, the quantizer 104 be configured to compress a range of values (eg, the amplitudes of the frequency components) to a single quantum value. That way, quantizers can be used 104 be configured to compress the data by presenting DCT coefficients with no greater precision than necessary to achieve the desired image quality. Therefore, quantizers 104 a sequence of quantized DCT coefficients (DC and AC) as input to entropy coders 106 deliver.

entropy 106 may be configured to perform the final DCT-based processing operations. entropy 106 can achieve additional compression lossless by more compact coding of the quantized DCT coefficients based on their statistical characteristics. entropy 106 may be configured to perform a variety of different operations, as discussed in greater detail below.

in the Generally, entropy coding can be two different operations exhibit. The first operation can quantize a zigzag sequence Convert coefficients into a cutscene of symbols. The purpose and operation of the zigzag sequence becomes more detailed in the "JPEG Still Picture Compression Standard", which was cited above, be written. The second operation can be the Convert symbols to a data stream where the symbols are none from outside identifiable borders have more.

Therefore, in some embodiments, encoders 106 convert the zigzag sequence of quantized coefficients into an intermediate sequence of signals. encoder 106 can also convert the symbols to a data stream where the symbols no longer have any externally identifiable borders. The form and definition of the intermediate symbols may depend on both the DCT based mode of operation and the entropy coding method (Huffman, arithmetic, etc.).

As discussed above, the input of the entropy coder 106 a sequence of quantized DCT coefficients (DC and AC). In the intermediate symbol sequence, each non-zero AC coefficient can be represented in combination with its "runlength". As used herein, the term "runlength" or "RUNLENGTH" may correspond to the sequence number of zero-valued AC coefficients preceding the non-zero AC coefficient in the zigzag sequence.

each Combination of run length / non-zero coefficient can be achieved by a pair of symbols are displayed, for example symbol-1 (RUNLENGTH, SIZE) and Symbol-2 (AMPLITUDE). In this example, symbol-1 represent the two pieces of information RUNLENGTH and SIZE. As mentioned above, RUNLENTH can be the number of consecutive zero-valued AC coefficients in the zigzag sequence, which precedes the non-zero coefficient shown. The term "SIZE" can as used herein, correspond to the number of bits used to encode AMPLITUDE (ie symbol-2). The term "AMPLITUDE" can As used herein, the amplitude of the non-zero AC coefficient correspond. Symbol 2 can be a single piece of information, which is referred to as AMPLITUDE represent.

in the Generally, the RUNLENGTH value can be zero runs of length represent zero to fifteen. Actual zero runs in the zigzag sequence can get bigger be as 15, so that the symbol 1 value (15, 0) as the extension symbol can be interpreted with RUNLENGTH = 16. The special symbol 1 value (0, 0) shows a situation at the end of the block (EOB) and can be considered as an "escape" symbol, which completes the 8x8 sample block.

Therefore can calculate the zigzag sequence 63 of quantized AC coefficients for every 8x8 block of samples as a sequence of symbol-1 and symbol 2 symbol pairs. However, every "couple" in the Case of a long RUNLENGTH repeat of Symbol-1 or just a symbol-1 in the case of an EOB.

Of the possible range of quantized AC coefficients may be the Determine the range of values to which both the AMPLITUDE and the SIZE information. For AC coefficients becomes a possible structure of the Symbol 1 and Symbol 2 intermediate representations Illustrated below in Tables 1 and 2, respectively. For DC coefficients the intermediate representation can be structured similarly. In this example, symbol-1 can only represent SIZE information; Symbol 2 can display AMPLITUDE information as before: Symbol 1 (SIZE), Symbol 2 (AMPLITUDE).

Because the DC coefficient can be differentially encoded, it can be covered by twice as many integer values [-2 ¹¹ , 2 ¹¹ -1] as AC coefficients, so adding an extra level at the end of Table 2 for DC coefficients can. Symbol 1 for DC coefficients may therefore represent a value from 1 to 11. SIZE 0 1 2 ... 9 10 RUNLENGTH 0 EOB RUNLENGTH VALUES , X , X , X 15 ZRL Table 1. Basic Huffman coding symbol 1 structure SIZE AMPLITUDE 1 1-, 1 2 -3, -2, 2, 3 3 -7 .. -4, 4 .. 7 4 -15 .. -8, 8 .. 15 5 -31 .. -16, 16 .. 31 6 -63 .. -32, 32 .. 63 7 -127 .. -64, 64 .. 127 8th -255 .. -128, 128 .. 255 9 -511 .. -256, 256 .. 511 10 -1023 .. -512, 512 .. 1023 Table 2. Basic entropy coding symbol 2 structure

Huffman coding is a method that uses symbols (eg bytes, DCT coefficients, etc.) and these with variable length codes can code according to statistical probabilities can be assigned. For example, the Huffman coding scheme, used with JPEG compression, file size by replacing the fixed-size codes (for example 12 bits) by variable-length codes (e.g. 1-16 Bits). On average, the replacement codes add up to fewer bits than the original. If on average 10% less Bits are needed to get the quantized DCT coefficients This can directly add to a 10% file size reduction to lead. The hard part may be deciding which codes to replace with which bit strings of variable length are. The Huffman tables define these figures and will stored within each JPEG image file. A separate Huffman table may be needed for each of the following four components are: luminance DC component, luminance AC component, Chrominance DC component and chrominance AC components.

For example These tables can be a table for luminance including AC coefficients (162 datasets, 162 × 20 = 3240 bits), a table for chrominance including AC coefficients (162 datasets, 162 × 20 = 3240 bits), a table of luminance including DCT coefficients (12 records, 12 × 15 = 180 bits) and a table for chroma including DC coefficients (12 records, 12 × 15 = 180 bits) exhibit. This can result in a total of 6840 bits or 855 bytes for standard entropy coding, as shown in Table 6.

• 1. Aktion-1: Für jedes Element von 63 AC-Koeffizienten kann diese mit Null verglichen werden. Daher können für jede Domäne 63 Aktion-1 vorliegen.
• 2. Aktion-2: Falls das vorliegende Element der Sequenz 0 ist, kann der Zähler der RUNLENGTH erhöht werden.
• 3. Aktion-3: Falls die Anzahl von RUNLENGTH größer als 15 ist und kein EOB vorliegt, kann der Zähler der RUNLENGTH modifiziert werden.
• 4. Aktion-4: Falls das vorliegende Element nicht null ist, kann eine Adresse für die passende Tabelle erzeugt werden, das Codewort kann erhalten und geschrieben werden.
• 5. Aktion-5: Sobald das Codewort für die RUNLENGTH-SIZE-Kombination erhalten wurde, kann der AMPLITUDEN-Wert in Binärform dargestellt und geschrieben werden.
• 6. Aktion-6: Adressenberechnung in Huffman-Tabellen (unter Verwendung von Werten von RUNLENGTH und SIZE).

Standard Run Length Encoding (RLE) - Data compression can have multiple operations. For example, the following operations may be implemented while generating an RLE sequence:

• 1. Action-1: For each element of 63 AC coefficients, this can be compared to zero. Therefore, there can be 63 action-1 for each domain.
• 2. Action-2: If the present element of the sequence is 0, the counter of the RUNLENGTH can be increased.
• 3. Action-3: If the number of RUNLENGTH is greater than 15 and there is no EOB, the counter of the RUNLENGTH be modified.
• 4. Action-4: If the present element is not zero, an address for the matching table can be generated, the codeword can be obtained and written.
• 5. Action-5: Once the code word for the RUNLENGTH-SIZE combination has been obtained, the AMPLITUDE value can be displayed in binary form and written.
• 6. Action-6: Address calculation in Huffman tables (using values of RUNLENGTH and SIZE).

Table 3 shows the average complexity values per block for the operations listed above using the YUV model, where Y is brightness (ie, luminance) and UV color (ie, chrominance). For example, if average action 3-Y complexity equals 0, 1, there may be only one action-3-Y for an average of ten blocks. Complexity of standard RLE Action-1-Y 63 Action-1 UV 63 Action 2 Y 48 Action 2 UV 57 Action 3 Y 0.1 Action 3 UV 0.1 Action-4-Y 16 Action-4-UV 7 Action-5-Y 15 Action-5 UV 6 Action-6-Y 16 Action-6 UV 7 Table 3. Complexity of standard RLE

Run Length Encoding with short tables

3 shows a system 300 for low complexity coding according to an exemplary embodiment of the present disclosure. system 3 can be configured to receive raw video data and this data using converters 302 convert from a red-green-blue (RGB) color space to a luminance and chrominance (YUV) color space.

Conversion from RGB to YUV can be accomplished using a variety of techniques, including but not limited to, the use of fast, table-driven algorithms. Each converted Y, U and V frame can be grouped in 8x8 blocks and in FDCT 304 be entered. Each 8x8 block of source image examples is effectively a 64-point discrete signal, which may be a function of x and y. FDCT 304 can receive this signal as input and decompose it into 64 orthogonal base signals. Each may include one of the 64 unique two-dimensional (2D) "spatial frequencies" that comprise the "spectrum" of the input signal. The output of FDCT 304 may be the set of 64 base signal amplitudes or "DCT coefficients" whose values can be uniquely determined by the particular 64-point input signal. FDCT can output this output to quantizers 306 deliver. It should be noted that every section of system 300 For example, it may be included in an integrated circuit or similar device.

After the issue of FDCT 304 For example, each of the 64 DCT coefficients may be uniformly quantized in conjunction with a quantization table (e.g., 64 elements) provided by the application (or user) as input to the entropy coder 308 can be specified. Each element may be an integer value from 1 to 255, which specifies the step size of the quantizer for its corresponding DCT coefficient. Quantization can be defined as the division of each DCT coefficient by its corresponding quantizer step size, followed by rounding to the nearest integer. This output value can then be normalized by the quantizer step size. After quantization, each element can be an entropy encoder 308 be entered.

entropy 308 can RLE encoding module 310 and Huffman coding module 312 lock in. The standard run-length encoding described above with reference to Table 3 requires four Huffman tables. On the other hand, in this embodiment, separate tables can be used for the RUNNING and SIZE values for AC coefficients. In this embodiment, the same tables can be used without changes for the DC coefficients. This can reduce the amount of memory needed to maintain the Huffman tables. In example, according to this off A table of codewords for all possible RUNLENGTH values consists of 16 data sets (16 × 10 = 160 bits for the UV components, 14 × 16 = 224 bits for the Y component), and the table for SIZE can contain 12 data records ( 12 × 16 = 192 bits). This can result in a total of 576 bits or 72 bytes. Therefore, this approach can result in a significant saving of memory resources compared to existing coding techniques, which requires 855 bytes.

Table 4 shows a summary showing the reduced number of operations needed when using separate tables for LAUFLÄNGLÄNGE and SIZE for AC coefficients. Complexity of RLE with short tables Action-1-Y 63 Action-1 UV 63 Action 2 Y 48 Action 2 UV 57 Action 3 Y 0.1 Action 3 UV 0.1 Action-4-Y 32 Action-4-UV 14 Action-5-Y 15 Action-5 UV 6 Table 4. Complexity of RLE with short tables

One measure of the complexity of this approach to RLE coding may include determining the number of operations required. The number of operations can be compared to that needed for standard RLEs, as shown in Table 3 above. The primary difference appears in action-4, because in this embodiment two different tables must be addressed, therefore the number of these operations is doubled. In addition, Action-6 is not needed to count addresses in Huffman tables. A more detailed comparison between standard RLE and this embodiment can be made by comparing 7 and 8th formed in more detail below.

4 shows a diagram 400 showing the rate-distortion curve (ie, compression rate versus quality of a reconstructed image using a PSNR metric) associated with standard JPEG entropy coding as compared to the entropy coding described in accordance with this embodiment of the present disclosure. The approach described above in 3 can simplify the encoding process, although the compression ratio may be slightly reduced (see 4 ), because RUNLENGTH and SIZE can be independently coded without using the relation between them. 4 shows that the compression rate and corresponding quality level (calculated using the rate distortion function) decreases very slightly. For example, the decrease in quality level for the embodiments described herein for rate = 1 Mbps is less than 0.1 dB compared to standard JPEG using a PSNR metric.

Run Length Encoding without tables for RUNLENGTH

In In some embodiments, run-length coding may be used performed without using Huffman tables for RUNLENGTH become. That is, for AC coefficients only Huffman tables can be used for SIZE. For DC coefficients can be the same tables as described above. Codewords for RUNLENGTH are not needed, instead you can these zeros are written to the output stream. For example can use the zeros if six zeros and non-zero AC coefficients in sequence, written to the stream, followed from SIZE and AMPLITUDE.

The determination to use only Huffman tables for SIZE may, after consideration of 5 and 6 which show the probability distribution of both RUNLENGTH and SIZE values, respectively. As in 5 As shown, the likelihood of a large zero sequence may be very small to test different test images (e.g., claire80.avi versus soldiers_hi.avi). Here the distribution function for the different test images does not seem to change significantly. For example, the probability of five consecutive zeros occurring in a sequence of AC coefficients may be less than 0.008 for the various test images.

In contrast, shows 6 the probability distribution for SIZE values. It can be seen here that the distribution function for SIZE values for different films may change and therefore be more complicated can. For example, the probability of SIZE 2 varies by approximately 0.45 for the claire80.avi to approximately 0.55 for the solders_hi.avi.

Table 5 shows the complexity in terms of the number of operations needed when RUNLENGTH encoding is used without the use of Huffman tables for RUNLENGTH. Eliminating the RUNLENGTH tables can also eliminate the need for the RUNLENGTH counter. Therefore, Action-2 and Action-3 (ie RUNLENGTH counter modifications) are not necessary. The Action 6 requirement can also be eliminated as there is no need to calculate addresses in the Huffman tables. Complexity of RLE without tables for run Action-1-Y 63 Action-1 UV 63 Action-4-Y 16 Action-4-UV 7 Action-5-Y 15 Action-5 UV 6 Table 5. Complexity of RLE without tables for run

Therefore, by eliminating the Huffman tables for RUNLENGTH, the process of coding can be simplified. As a result, the compression ratio can also, more than in 4 is decreasing because the RUNLENGTH values are not compressed.

7 to 9 Figure 12 shows a series of flowcharts comparing standard JPEG entropy coding with the embodiments of the present disclosure. For example, shows 7 a flowchart 700 including operations 701 to 715 which are used to perform standard entropy coding according to the JPEG image compression algorithm. flow chart 700 includes the operation of address calculation in the Huffman tables for RUNLENGTH and SIZE (ie action-6). The standard entropy coding technique used in 7 shown may require 855 bytes of memory usage.

In contrast, shows 8th flow chart 800 including operations 801 to 814 configured to perform entropy coding using separate tables for RUNLENGTH and SIZE according to the embodiments described herein. In this embodiment, as discussed above, the operation of address calculation in Huffman tables is removed (therefore no action-6 is needed). In the 8th The embodiment shown may also require much less memory (ie 72 bytes) than standard entropy coding.

9 shows a flowchart 900 including operations 901 to 911 configured to perform entropy coding in accordance with the embodiments described herein. flow chart 900 FIG. 12 shows an embodiment for performing entropy coding without using tables for RUNLENGTH. In this embodiment, the operations related to the calculation of RUNLENGTH may be cleared. For example, operations are 705 (ie action-2, which equals 1 to increment the run counter) 707 . 709 and 710 of the 7 deleted. flow chart 900 also shows the writing of zero coefficients into the stream as in operation 905 is shown. In the 9 The embodiment shown may also require less memory (ie 24 bytes) than standard entropy coding.

A summary showing the standard entropy coding as well as the embodiments described herein is given in Table 6 below. Table 6 shows the memory requirement variations between (1) existing entropy coding techniques, (2) entropy coding using short RL tables, and (3) entropy coding without using tables for RUNLENGTH. As shown below, the embodiments described herein provide a significant amount of memory savings compared to standard entropy coding. Entropiekodierungsschema operational complexity memory consumption Entropy coding - standard (1) 126 Action-1 +105 Action-2 +0,2 Action-3 +23 Action-4 +21 Action-5 +23 Action-6 855 bytes Entropy coding - short RL tables (2) 126 action-1 +105 action-2 +0,2 action-3 +46 action-4 +21 action-5 72 bytes Entropy coding - RLE without tables for RL (3) 126 action-1 +23 action-4 +21 action-5 24 bytes Table 6. Memory consumption for the three techniques

According to an exemplary embodiment of the present disclosure, in 10 a flowchart 1000 provided by data compression operations. The method may include converting data from a first color space to a second color space to convert converted data (Fig. 1002 ) to create. The method may further comprise receiving the converted data at a Discrete Cosine Transform (DCT), wherein the DCT is configured to generate at least one DCT coefficient (DCT). 1004 ) to create. The method may additionally comprise receiving the at least one DCT coefficient from the discrete cosine transform at a quantizer, wherein the quantizer is configured to quantize the at least one DCT coefficient and provide at least one quantized coefficient to an entropy coder, the entropy coder having a first Huffman table corresponding to a number of bits used to calculate an amplitude of a non-zero AC coefficient ( 106 ). Of course, additional operations are also within the scope of the present disclosure.

The Embodiments described herein can be used for example in an integrated circuit (integrated circuit (IC)), for example, a system-on-a-chip (SoC), an application-specific integrated circuit (application specific integrated circuit (ASIC)) and / or a field programmable gate array (FPGA). "Integrated circuit", as used in each embodiment herein a semiconductor device and / or microelectronics device, such as, for example, a semiconductor integrated circuit chip, without being limited to these.

As in each embodiment described herein For example, "circuits" can be used individually or in any combination hard-wired circuits, programmable circuits, state machine circuits and / or Include firmware that store commands by programmable Circuits are executed. It's easy to see that any of the operations and / or operational components that are in each embodiment is described herein in software, Firmware, hardwired circuits, and / or any combination the same can be implemented.

embodiments The methods described above can be used in a computer program be implemented, which are stored on a storage medium may be the commands for programming a system (for example a machine) for carrying out the methods. The storage medium can include any type of disk Floppy Disks, Optical Discs, Compact Disk Read Only Memory (CD-ROMS), Compact Disk Rewritables (CD-RWs) and Magneto-Optical Disks, Semiconductor Devices, such as read-only memory (ROMs), random access memory (RAMs), such as dynamic and static RAM, erasable programmable Read-only memories (EPROMs), electrically erasable programmable read-only memories (EEPROMs), flash memory, magnetic or optical cards or any type of media suitable for storing electronic commands are, have. Other embodiments may be considered Software modules by a programmable controller to be implemented.

The present disclosure can provide numerous advantages over the prior art. at For example, the embodiments described herein describe a number of low complexity entropy encoding schemes that reduce the computational complexity of VLC-based entropy encoders by approximately 50% and memory consumption by approximately 70% compared to standard JPEG. The embodiments described herein allow selection of an efficient trade-off between computational complexity and compression ratio.

For example, the embodiments described herein need not require coding the null sequences between non-zero coefficients (RUNLENGTH). In addition, in some embodiments, Huffman tables are not needed for RUNLENGTH values, so memory consumption (see Table 6) is reduced. It has been shown that the probability of such sequences for different test films (see 5 ) can be very small and that the effect of the compression ratio is very small (see 4 ).

numerous Features, aspects, and embodiments have been described herein. The features, aspects and embodiments are subject a combination with each other, as well as variation and modification, as will be apparent to those skilled in the art. The present disclosure should Therefore, such combinations, variations and modifications are considered to include.

Summary

The The present disclosure provides a method of data compression ready. In one embodiment, the method may convert of data from a first color space to a second color space, to generate converted data. The method may further receive the converted data at a discrete cosine transform (DCT), wherein the DCT is configured to be at least one To produce DCT coefficients. The procedure may additionally Receiving the at least one DCT coefficient from the discrete one Cosine transform at a quantizer, wherein the quantizer is configured to quantize the at least one DCT coefficient and at least one quantized coefficient for one Entropiekodierer provide, wherein the Entropiekoder a has first Huffman table corresponding to a number of bits, which is used to produce an amplitude of a non-zero AC coefficient to code.

QUOTES INCLUDE IN THE DESCRIPTION

This list The documents listed by the applicant have been automated generated and is solely for better information recorded by the reader. The list is not part of the German Patent or utility model application. The DPMA takes over no liability for any errors or omissions.

Cited non-patent literature

• Joint Photographic Experts Group (JPEG) standard, ITU-T T.81 [0017]
• - ISO / IEC IS 10918-1 [0017]
• - "The JPEG Still Picture Compression Standard", Gregory K. Wallace, IEEE Transactions on Consumer Electronics, 1991 [0019]
• - "The JPEG Still Picture Compression Standard", Gregory K. Wallace, IEEE Transactions to Consumer Eletronics, 1991 [0021]

Claims (15)

Apparatus comprising: an integrated An integrated circuit (IC) configured to Data to compress and / or decompress, the integrated Circuit has an entropy coder configured at least one discrete cosine transform (DCT) coefficient the entropy coder is a first Huffman table which corresponds to a number of bits that is used to encode an amplitude of a non-zero AC coefficient. Device according to claim 1, characterized in that the entropy coder further comprises a second Huffman table, that of a consecutive number of zero-valued AC coefficients corresponds to the non-zero AC coefficient in a given Sequence precedes. Device according to claim 1, characterized in that the entropy coder is configured to be at least one zero to write to an output stream, the at least one Zero of a consecutive number of zero-valued AC coefficients corresponds to the non-zero AC coefficient in a given Sequence precedes. The device of claim 1, further comprising Quantizer configured to have the at least one coefficient receive from a forward discrete cosine transformation, wherein the quantizer is further configured to receive the at least to quantize a coefficient and at least one quantized one To provide coefficients to the entropy encoder. Device according to claim 4, characterized in that that the forward discrete cosine transformation is configured to receive converted data from a converter, where the discrete cosine transform is further configured to execute the to generate at least one DCT coefficient and the at least to provide a DCT coefficient to the quantizer. A method of data compression, comprising: Convert from data from a first color space to a second color space to generate converted data; Receive the converted Data at a Discrete Cosine Transform (DCT), where the DCT is configured to generate at least one DCT coefficient; and Receiving the at least one DCT coefficient from the discrete cosine transform at a quantizer, wherein the Quantizer is configured to the at least one DCT coefficient to quantize and at least one quantized coefficient to an entropy coder, the entropy coder has a first Huffman table that corresponds to a number of bits, which are used to calculate an amplitude of a non-zero AC coefficient to code. Device according to claim 6, characterized in that the entropy coder further comprises a second Huffman table, that of a consecutive number of zero-valued AC coefficients corresponds to the non-zero AC coefficient in a given Sequence precedes. Method according to Claim 6, characterized the entropy coder is configured to be at least one zero to write to an output stream, the at least one Zero of a consecutive number of zero-valued AC coefficients corresponds to the non-zero AC coefficient in a given Sequence precedes. The method of claim 7, further comprising: to compare the at least one DCT coefficient of zero; Increase of a counter which zeroes out of a consecutive number AC coefficient corresponding to the non-zero AC coefficient precedes in a given sequence; Modifying the counter, if the consecutive number of zero-valued AC coefficients, which precedes the non-zero AC coefficient in a given sequence, is greater than a given threshold and no End of block (EOB) is; Detecting a non-zero coefficient and obtaining and writing a codeword that is at least one successive one Number of zero-valued AC coefficients corresponding to the non-zero AC coefficient precedes in a given sequence corresponds, and wherein the Number of bits used is an amplitude of a non-zero AC coefficient to code; and Write the amplitude of the non-zero AC coefficient in binary form. The method of claim 8, further comprising: to compare the at least one DCT coefficient of zero; detect a non-zero coefficient and getting and writing a codeword, the at least one consecutive number of zero-valued AC coefficients corresponds to the non-zero AC coefficient in a given Sequence precedes, and where the number of bits is used to encode an amplitude of a non-zero AC coefficient; and Write the amplitude of the non-zero AC coefficient in binary form. Product comprising a storage medium thereon has stored instructions when executed by lead a machine to: Convert from Data from a first color space to a second color space converted to Generate data; Receive the converted data on one discrete cosine transform (DCT), where the DCT is configured is to produce at least one DCT coefficient; and Receive of the at least one DCT coefficient from the discrete cosine transform at a quantizer, where the quantizer is configured, to quantize the at least one DCT coefficient and at least provide a quantized coefficient to an entropy encoder, the entropy coder having a first Huffman table, which corresponds to a number of bits used to make a Amplitude of a non-zero AC coefficient to code. Product according to claim 11, characterized in that the entropy coder further comprises a second Huffman table, that of a consecutive number of zero-valued AC coefficients corresponds to the non-zero AC coefficient in a given Sequence precedes. Product according to claim 11, characterized in that the entropy coder is configured to be at least one zero to write to an output stream, the at least one Zero of a consecutive number of zero-valued AC coefficients corresponds to the non-zero AC coefficient in a given Sequence ahead. The article of claim 12, further comprising: to compare the at least one DCT coefficient of zero; Increase of a counter which zeroes out of a consecutive number AC coefficient corresponding to the non-zero AC coefficient precedes in a given sequence; Modifying the counter, if the consecutive number of zero-valued AC coefficients, which precedes the non-zero AC coefficient in a given sequence, is greater than a given threshold, and is not an end-of-block (EOB); detect a non-zero coefficient and getting and writing a codeword, the at least one of a consecutive number of zeros AC coefficient corresponding to the non-zero AC coefficient in a given sequence and where the number of bits is used to calculate an amplitude of a non-zero AC coefficient to code; and Write the amplitude of the non-zero AC coefficient in binary form. The article of claim 13, further comprising: to compare the at least one DCT coefficient of zero; detect a non-zero coefficient and getting and writing a codeword, the at least one of the consecutive number of zeros AC coefficient corresponding to the non-zero AC coefficient precede in a given sequence, and where the number of Bits is used to give an amplitude of a non-zero AC coefficient to code; and Write the amplitude of the non-zero AC coefficient in binary form.