JP3532440B2 - Cosine transform and quantization device, multiplicative operation saving method of image compression device, inverse quantization and inverse cosine transform device, and multiplicative operation saving method of image decompression device - Google Patents

Description

Detailed Description of the Invention

[0001]

BACKGROUND OF THE INVENTION 1. Field of the Invention The present invention relates to an image compression and image decompression technique, and more particularly to an image compression device and an image decompression device capable of greatly reducing the number of multiplication operations.

[0002]

2. Description of the Related Art Discrete cosine transform (DCT) is widely used for image compression operations such as MPEG for dynamic image coding and JPEG for static image coding. Similarly, inverse discrete cosine transform (IDCT) is used. ) Is used for the inverse process of decompressing the compressed image data corresponding to the above.

DCT and IDCT operations typically have to involve multiple multiplications and additions, and usually the DCT and IDCT processing multiplications are time consuming and require fairly complex hardware. US Patent 54714
In No. 12, the applicant has a DCT / IDC of six-step calculation
The T-speed calculation method is disclosed, which includes only 13 multiplication operations for one-dimensional conversion, and only a total of 208 times (2 × 8 × 1) for two-dimensional conversion of 8 × 8 data blocks.
3) Multiplication operation is included. 7 and 8 show the DCT
And IDCT speed calculation method, it is desirable to further reduce the number of multiplication operations in order to obtain a higher processing speed in image compression and / or image decompression.

[0004]

SUMMARY OF THE INVENTION Therefore, a first object of the present invention is to provide a cosine transform and quantizer used in an image compression apparatus, which requires a relatively small number of multiplication operations compared with the prior art. Therefore, a second object of the present invention is to provide a multiplicative arithmetic operation saving method for an image compression apparatus, which can expect a higher processing speed. A fourth object of the present invention is to provide an inverse quantization and inverse cosine transform device used for image compression and image decompression, which requires a relatively small number of executions of operations. An object of the present invention is to provide a method for reducing the multiplication operation of a decompression device.

[0005]

In order to achieve the first object, the cosine transform and quantizer used in the image compression apparatus of the present invention includes a discrete cosine transform unit and a quantizer. The discrete cosine transform unit performs a discrete cosine transform (DCT) on the input data block based on the DCT speed calculation method of the six-step operation, and the discrete cosine transform unit omits the sixth step internal multiplication operation of the DCT speed calculation method. To obtain the transformed data after the scaling transformation, the quantizer is connected to the discrete cosine transformation unit to quantize the transformed data after the scaling transformation based on the modified quantizer matrix, The quantized DCT data corresponding to the data is obtained.
Then, the modified quantizer matrix is obtained by compensating a predetermined primitive quantizer matrix with a scaling coefficient derived from the sixth stage internal multiplication operation of the DCT speed calculation method.

In order to achieve the above second object, a method of multiplicative arithmetic reduction of an image compression apparatus according to the present invention comprises a discrete cosine transform unit and a quantizer connected to the discrete cosine transform unit. The multiplicative arithmetic reduction method comprises a step of compensating a predetermined primitive quantizer matrix with a scaling factor derived from a sixth stage internal multiplicative operation of the six-stage operation discrete cosine (DCT) speed calculation method to obtain a modified quantizer matrix. And, the discrete cosine transform unit is provided under the condition that the sixth step internal multiplication operation of the DCT speed calculation method is omitted, the input data block is subjected to DCT by the DCT speed calculation method, and the converted data that has undergone the scaling conversion is obtained. And a quantizer to quantize the transformed data by the modified quantizer matrix to obtain the input data block. Tsu comprising obtaining a quantized DCT data corresponding to the click, the.

In order to achieve the third object, the inverse quantization and inverse cosine transform apparatus used for image compression and image reduction of the present invention includes an inverse quantizer and an inverse discrete cosine transform unit. The quantizer dequantizes the input quantized discrete cosine transform (IDCT) data by the modified solution quantizer matrix to obtain the solution quantized data that has undergone scaling transformation, and the modified solution quantizer matrix is a six-step operation inverse discrete It is obtained by compensating a predetermined primitive solution quantizer matrix by the prescaling coefficient derived from the first stage internal multiplication operation of the cosine (IDCT) speed calculation method. The inverse discrete cosine transform unit is connected to the inverse quantizer, performs IDCT on the solution quantized data that has undergone the scaling transformation by the six-step operation IDCT speed calculation method, and the inverse discrete cosine transform unit performs the IDCT. The IDCT data corresponding to the input quantized DCT data can be obtained by omitting the first stage internal multiplication operation of the speed calculation method.

In order to achieve the above-mentioned fourth object, according to the method of multiplicative arithmetic reduction of an image decompressor of the present invention, the image decompressor comprises an inverse quantizer and an inverse discrete cosine transform unit connected to the inverse quantizer. The multiplicative operation saving method includes a modified solution quantization by compensating a predetermined primitive solution quantizer matrix with a prescaling coefficient derived from the first-stage internal multiplication operation of the six-step operation inverse discrete cosine (IDCT) speed calculation method. A quantizer matrix, a step of providing an inverse quantizer, dequantizing input quantized discrete cosine transform (DCT) data by the modified solution quantizer matrix, and obtaining solution quantized data after scaling transformation. , The first-stage internal multiplication operation of the IDCT speed calculation method is omitted, the inverse discrete cosine transform unit is provided, and the solution quantization through the scaling transform is performed. Performs IDCT on over data, other features and advantages of the present invention comprising the steps of obtaining an IDCT data corresponding to the input quantized DCT data, the it can be understood from the following detailed description while referring to the drawings.

[0009]

BEST MODE FOR CARRYING OUT THE INVENTION The present invention will be specifically described below based on the embodiments, but the present invention is not limited to this example. FIG. 1 is a circuit display block diagram of an image coding apparatus for discrete cosine transform, in which a motion estimator 10 receives a source input frame and performs a motion estimation related to a pre- and / or post-input frame, Is a predictive coded frame or a bidirectional predictive coded frame, the motion estimator 10 produces a motion vector, and both the predictive coded frame and the bidirectional predictive coded frame are generally non-intra coded frames. Or, it is called an intra coded frame.
The input frame is directly supplied to the cosine transform and quantizer 20, and the quantizer 20 uses the discrete cosine transform (DC
The T) unit 21 and the quantizer (Q) 22 connected to the DCT unit 21 are included. The DCT unit 21 performs a discrete cosine transform on the input frame, and provides the transformed data to the quantizer 22 so that the quantizer 22 can transform the transformed data to a finite level from some amplitude in a preset range. To quantize. The non-intrinsic coded frames are also processed by the cosine transform and quantization device 20,
However, it is only after the reference pixel value from the reference frame stored in the reference frame buffer 30 is subtracted from the pixel value of the current frame by the subtractor 11. Thus, within the non-intrinsic coded frame, the cosine transform and quantizer 20 only processes the difference between those pixel values. The inverse quantizer and inverse cosine transform device 40 includes an inverse quantizer (De
Q) 41 and an inverse discrete cosine transform (IDCT) unit 42 connected to the inverse quantizer 41.
Performs the inverse operation of the quantizer 22, and the IDCT unit 4
2 performs the inverse operation of the DCT unit 21. The reference frames of the non-intrinsic coded frames are supplied from the adder 12, and the reference frames may be intrinsic coded frames, which are directly constructed after being newly constructed by the inverse quantization and inverse cosine transform device 40. It is stored in the frame buffer 30. Alternatively, the reference frame may be a predictive coded frame, and is constructed one step further by referring to the advance reference frame stored in the reference frame buffer 30.
In this situation, the adder 12 compensates the difference between the pixel values of the reference frame of the predictive coding frame after being recovered through the IDCT of the predictive coding frame, and the newly constructed predictive coding frame to the reference frame buffer. Send to 30.
The motion vector from the motion estimator 10 is used to address the appropriate pixels of the reference frame in the reference frame buffer 30 to perform the subtraction operation to encode the non-intrinsic coded frame, Alignment compensation can be performed to newly construct a predictive coding reference frame. A variable length coder (VL) connected to the output terminal of the quantizer 22.
C) 13 is used for performing entropy coding such as Huffman coding, and associates the quantized transform data with the variable length code of the compressed image code that is continuously output.

The motion estimator 10 removes extra parts between the frames of the source input frame, the quantizer 22 removes unimportant parts of the image data which are not perceptible to the naked eye, and the variable length encoder 13 entropy. By applying the encoding technology, it is possible to go one step further and compress the image data to obtain compressed image data having a compression ratio range of several tens to several hundreds.

If the source input frames are static images, such as photographs, with no inter-frame correlation, the image coder of FIG. 7 is simply a cosine transform and quantizer 20 and a variable length code. It can be reduced to include only the quantizer 13, and includes a motion estimator 10, a subtractor 11 used to provide a pixel value difference, an adder 12 used to provide a reference frame, and a reference frame to be stored. The reference frame buffer 30 used and the inverse quantization and inverse cosine transform device 40 can be omitted.

The general concept of the image signal encoding process and the encoding device described above is a well-known conventional technique, which is widely used in digital image compression techniques such as MPEG and JPEG, and uses a cosine transform and a quantizer. US Pat. No. 5,748,240;
No. 5,321,725 and 4,245,248.

FIG. 2 is a circuit display block diagram of an image decoding device corresponding to the image encoding device of FIG. 1. The image decoding device is a variable length decoder (VLD) 50, an inverse quantization and inverse cosine transform device 60. , Reference frame buffer 70, and adder 8
The inverse quantization and inverse cosine transform device 60 includes 0 and includes an inverse quantizer (DeQ) 61 and an inverse discrete cosine transform (IDCT) unit 62 connected to the inverse quantizer 61. The variable length decoder 50 can process those compressed image codes that are variable length codes to obtain quantized DCT data. The inverse quantizer 61 performs an inverse quantization operation on the quantized data to obtain a solution quantized DCT. IDCT unit 6
2 performs IDCT on the dequantized DCT data from the inverse quantizer 61 to obtain IDCT data. The IDCT data is a pixel value in terms of an intrinsically encoded frame, and a pixel value difference from a reference frame in terms of a non-intrinsically encoded frame. The operation of the reference frame buffer 70 is the same as that of the reference frame buffer 30 in the image coding apparatus of FIG. The adder 80 compensates the IDCT data with the reference frame from the reference frame buffer 70 of the non-intrinsic coded frame, and the intrinsic coded frame from the IDCT unit 62 and the compensated non-intrinsic coded frame are newly constructed. Build up the frame. The newly constructed intrinsic coded frame and predictive coded frame are further stored in the reference frame buffer 70 to act as a reference frame. The reference frame buffer 70 has a motion vector input, which is used to address the appropriate pixels of the reference frame in the reference frame buffer 70 to compensate in newly constructing the non-intrinsic coded frame. Perform operational processing.

Similarly, if the compressed image codes are mere static images, such as photographs, which do not include inter-relationships between frames, the image decoding apparatus of FIG. It may be reduced to include only the encoding and inverse cosine transform unit 60, and the adder 80 and the reference frame buffer 70 used for newly constructing the non-internally encoded frame are omitted.

The general concept of the image signal decoding process and the decoder as described above belongs to the known prior art.
It is widely used in digital image reduction techniques such as MPEG and JPEG. In the cosine transform and quantizer, the quantization operation is usually performed after the discrete cosine transform operation. In the inverse quantization and inverse cosine transform device,
The inverse discrete cosine transform operation is usually performed after the inverse quantization operation.

The DCT / IDCT apparatus disclosed by the applicant in US Pat. No. 5,471,412 performs DCT on one 8 × 8 input data block based on the six-step arithmetic DCT speed calculation method shown in FIG. In addition, the IDCT is performed on one 8 × 8 input conversion data block based on the six-step operation IDCT speed calculation method shown in FIG. As shown in FIG. 7, the six-step operation of the DCT speed calculation method includes a first-step operation including one butterfly butterfly operation, a second-step operation including a multiplication operation after one two addition operations, and one A third stage operation with four butterfly operations, a fourth stage operation with a multiplication operation after one three additions, a fifth step operation with one four butterfly operations, and one eight Includes a sixth stage operation comprising an internal multiplication operation of. The two-dimensional DCT includes continuous first and second one-dimensional DCT operations, and the second one-dimensional DCT operation transforms the one-dimensional conversion data from the first one-dimensional DCT operation to correspond to the input data block. To obtain two-dimensional converted data.

As shown in FIG. 8, the six-step operation of the IDCT speed calculation method includes a first-step operation including one eight internal multiplication operations, a second-step operation including one four butterfly operations, and one-step operation. Third stage operation with subtraction operation after three synergistic operations, Fourth staircase operation with four butterfly operations, Fifth stage operation with subtractive operation after one two synergistic operations, and one It includes a sixth staircase operation with four butterfly operations. The two-dimensional IDCT includes continuous first and second one-dimensional IDCT operations, the second one-dimensional IDCT operation inversely transforms the one-dimensional inverse transform data from the first one-dimensional IDCT operation to obtain an input data block. The two-dimensional inverse transformation data corresponding to is obtained.

In the quantizer of the cosine transform and quantizer, the quantizing operation is performed on the 8 × 8 block reference using one predetermined quantizer matrix suitable for different image attributes. That is, the predetermined quantizer matrix varies depending on the luminance and chrominance components, the intrinsic coded frame and the non-intrinsic coded frame, or one image order sequence to another image order sequence. The equation is M (i, j) = f (i, j) / Q (i, j)
Then, among them, i, j = 0, 1, 2, ... 7, f (i,
j) represents DCT data, Q (i, j) represents a predetermined quantizer matrix, and M (i, j) represents quantized transform data.

In the inverse quantizer of the inverse quantization and inverse cosine transform device, the inverse quantization operation is also performed on the 8 × 8 block standard by using one predetermined solution quantizer matrix, and the solution quantization operation is performed. One basic equation is f ′ (i, j) =
M (i, j) × Q (i, j), of which i, j =
0, 1, 2, ..., 7, f ′ (i, j) represents the solution quantization transform data, Q (i, j) represents a predetermined solution quantizer matrix, and M (i, j) is Represents quantized transformed data.

Referring again to FIG. 1, unlike the prior art, the characteristic of the cosine transform and quantizer 20 of the preferred embodiment of the present invention is that the DCT unit 21 is a sixth stage internal multiplication operation of the DCT speed calculation method. Is omitted, DCT is performed on the input data block by the DCT speed calculation method,
Obtain converted data that has undergone scaling conversion. As a result, the quantizer is constructed so as to quantize the transformed data that has undergone the scaling transformation based on the modified quantizer matrix, and the quantized DCT data corresponding to the input data block can be obtained without any functional deterioration. The modified quantizer matrix is obtained by compensating a given primitive quantizer matrix with a scaling factor derived from a set of sixth-step internal multiplication operations of the DCT speed calculation method, and the overall processing speed of the cosine transform and quantizer 20 is obtained. Is greatly improved by this.

That is, by omitting the sixth stage internal multiplication operation of the DCT speed calculation method, the number of multiplication operations performed by the DCT unit 21 can be greatly reduced. Particularly, from the viewpoint of one-dimensional DCT transformation, the total number of multiplication operations is one 8 ×.
The number of eight input data blocks is reduced from thirteen to five, and in terms of the scaling transform one-dimensional DCT, it is sufficient to simply hold about 38% of multiplication operations. Then, the two-dimensional DCT transform requires only a total of 80 (2) to complete the scaling transform two-dimensional DCT of the 8 × 8 input data block.
Only the multiplication operation of x8x5) is required.

The transformed data that has undergone the scaling transformation from the DCT unit 21 of the relatively preferred embodiment of the present invention is two-dimensional transformed data. FIG. 3 is an operation flow chart of the cosine transform / quantization device 20. First, the DCT unit 21 performs the input data block D shown in FIG.
A one-dimensional conversion data that has undergone scaling conversion is obtained by performing a scaling conversion first-cycle one-dimensional conversion operation including operations of the first to fifth steps of the CT speed calculation method. Next, the DCT unit 21 performs a scaling conversion second cyclic one-dimensional conversion operation including the first to fifth steps of the DCT speed calculation method on the one-dimensional conversion data that has undergone the scaling conversion, and then performs the scaling conversion. Obtain dimension conversion data. Then, the quantizer 22 quantizes the two-dimensional transformed data that has undergone scaling transformation based on the modified quantizer matrix, and quantizes the two-dimensional DCT corresponding to the input data block.
Get the data block.

Modified quantizer matrix Q '(i, j)
Is obtained by the following equation. Q '(i, j) = Q (i, j) / C (i, j) where Q (i, j) represents a given primitive quantizer matrix and C (i, j) is its scaling Represents a set of coefficients. As shown in FIG. 4, the scaling coefficient set C (i, j) is the product of the sixth-stage weighting coefficients of the DCT speed calculation method during the first-round one-dimensional DCT transform and the second-round one-dimensional DCT transform.

Quantization equations may be associated with other items, such as rounding the sign consideration of the transformed data after scaling transformation, and merging the scaler into the quantizer matrix (eg MPEG quantizer). Scale). If companion items appear, the scaling factor set is populated before these items are quantized.

As a result, the DCT unit 21 and the DC
In order to reduce the multiplication operation of the image compression apparatus including the quantizer 22 connected to the T unit 21, the predetermined primitive quantizer matrix is derived from the sixth-step internal multiplication operation of the six-step discrete cosine (DCT) speed calculation method. A modified quantizer matrix is obtained by compensating the derived set of scaling factors. After that, the DC unit is constructed to perform the DCT speed calculation method on the input data block after removing the sixth-stage internal multiplication operation of the DCT speed calculation method, and the converted data that has undergone the scaling conversion can be obtained. The quantizer 22 quantizes the transformed data that has undergone the scaling transformation based on the modified quantizer matrix to obtain quantized data corresponding to the input data block.

As described above, the quantization operation takes a little longer in the operation time, but the DCT operation can be surprisingly shortened in the operation time. Therefore, when the cosine transform and the quantizer 20 are used in the image compression period, the whole effect is obtained. Can be improved. As shown in FIGS. 1 and 2, the inverse quantizer and inverse cosine transform devices 40 and 60 of the preferred embodiment of the present invention are characterized in that their inverse quantizers 41 and 61 are modified solution quantization matrices. Is used to dequantize the input quantized DCT data to obtain scaling-transform quantized data. Modified solution Quantization matrix is I
It is obtained by compensating a predetermined primitive solution quantization matrix by the prescaling coefficient set derived from the first stage internal multiplication operation of the DCT speed calculation method. After that, the first-stage internal multiplication operation of the IDCT speed calculation method is omitted, and the IDCT units 42 and 62 perform IDCT on the scaling conversion solution quantized data by the six-step operation IDCT speed calculation method, and input without deteriorating the function. IDCT data corresponding to the quantized DCT data can be obtained. As a result, the overall processing speed of the inverse quantization and inverse cosine transform devices 40 and 60 is greatly improved.

By omitting the first stage internal multiplication operation of the IDCT speed calculation method, the number of multiplication operations performed by the IDCT units 42 and 62 is greatly reduced. Especially one-dimensional IDC
From the T-transform, the total number of multiplication operations is reduced from 13 to 5 for each 8 × 8 input transform data block,
That is, from the viewpoint of the scaling conversion one-dimensional IDCT, it is sufficient to simply hold approximately 38% of the multiplication operation. The two-dimensional IDCT transform requires only a total of 80 multiplication operations (2 × 8 × 5) to complete the scaling transform two-dimensional DCT of the 8 × 8 input transform data block.

IDCT of a relatively preferred embodiment of the present invention
IDCT data from units 42 and 62 is a two-dimensional ID
It is CT conversion data. FIG. 5 is an operation flowchart of the inverse quantization and inverse cosine transform devices 40 and 60 shown in FIGS. 1 and 2. First, the inverse quantizers 41 and 61 modify the input quantized DCT data block and the modified solution quantizer matrix To obtain quantization data by scaling transformation. Next, the IDCT units 42 and 62 perform a scaling transformation first-round one-dimensional inverse transformation operation including the second to sixth steps of the IDCT speed calculation method on the scaling transformation solution quantized data to obtain the scaling transformation one-dimensional. Get the inverse transformed data. After that, the IDCT units 42, 62
Is the scaling transformation one-dimensional inverse transformation data I
The IDCT data corresponding to the input quantized DCT data block is obtained by performing the scaling transformation second-round one-dimensional inverse transformation calculation including the second to sixth stages of the DCT speed calculation method.

Modified Solution Quantizer Matrix Q '(i,
j) is obtained by the following equation. Q ′ (i, j) = Q (i, j) × P (i, j) where Q (i, j) represents a given primitive solution quantizer matrix, and P (i, j) is Represents a set of pre-scaling coefficients. As shown in FIG. 6, the prescaling coefficient set P (i, j) is the first cyclic one-dimensional ID.
It is the product of the first-stage weighting factors of the IDCT speed calculation method during the CT transformation and the second one-dimensional IDCT transformation.

Quantization equations can be associated with other items, such as quantization and adjustment of code consideration of the dequantized DCT transform data, and merging of the scaler into the quantizer matrix (eg, MPEG Quantizer scale). If companion items appear, the prescaling coefficient set is populated before these items are quantized.

As a result, the inverse quantizers 41 and 61 and the IDCT unit 4 connected to the inverse quantizers 41 and 61 are connected.
In order to reduce the multiplicative calculation of the image decompressor including 2,62, a given primitive solution quantizer matrix has a six-stage ID.
A modified solution quantizer matrix is obtained by using a set of prescaling coefficients derived from the first stage internal multiplication operation of the CT speed calculation method as compensation. After that, the inverse quantizers 41 and 6
1 is constructed to dequantize the input quantized DCT data based on the modified solution quantizer matrix to obtain scaling transformed solution quantized data. Then, the IDCT units 42 and 62 perform IDCT on the scaling conversion solution quantized data by the IDCT speed calculation method while omitting the first-stage internal multiplication operation of the IDCT speed calculation method, and the IDCT corresponding to the input quantized DCT data. Built to get the data.

As described above, the inverse quantization operation takes a little longer in the operation time, but the IDCT operation can be surprisingly shortened in the operation time. Therefore, when the inverse quantization and inverse cosine transform devices 40 and 60 are used in the image decompression period. Can improve the overall effect

[0033]

According to the present invention configured as described above, a cosine transform and a quantizer used in an image compression apparatus in which the number of multiplication operations executed is relatively smaller than that in the prior art, and a higher processing speed is achieved. The expected multiplicative operation saving method of the image compression apparatus, and the inverse quantization and inverse cosine transform apparatus used for image compression and image decompression, in which the number of executions of the multiplication operation is relatively small compared to the prior art, and higher processing speed are achieved. It is possible to provide a promising method for reducing the multiplication operation of the image decompression device.

[Brief description of drawings]

FIG. 1 is a block diagram showing an electric circuit of a relatively preferable embodiment of an image coding apparatus in which a cosine transform / quantization apparatus and an inverse quantization / inverse cosine transformation apparatus of the present invention are combined.

FIG. 2 is an electrical circuit block diagram of a relatively preferred embodiment of the image decompression device incorporating the inverse quantization and inverse cosine transform device of the present invention.

FIG. 3 is an operational flowchart of a cosine transform and quantizer according to a relatively preferred embodiment of the present invention.

FIG. 4 is a diagrammatic representation of scaling transform coefficient sets used during quantizer matrix modification of a quantizer in a relatively preferred embodiment of the present invention.

FIG. 5 is an operation flowchart of the inverse quantization and inverse cosine transform device in a relatively preferred embodiment of the present invention.

FIG. 6 is a diagrammatic representation of a set of pre-scaling transform coefficients used in modifying the solution quantization matrix of the dequantizer in a relatively preferred embodiment of the present invention.

FIG. 7 is a flow graph of a six-step operation DCT speed calculation method used in a one-dimensional conversion operation of a conventional DCT unit.

FIG. 8 is a flow graph of a six-step operation IDCT speed calculation method used for a one-dimensional inverse transform operation of a conventional IDCT unit.

[Explanation of symbols]

10 Motion estimator 20 Cosine transform and quantizer 21 Discrete Cosine Transform Unit 22 Quantizer 30 reference frame buffer 11 Subtractor 40 Inverse quantization and inverse cosine transform device 41 Dequantizer 42 Inverse Discrete Cosine Transform Unit 12 adder 13 Variable length encoder 50 variable length decoder 60 Inverse quantization and inverse DCT device 70 reference frame buffer

   ─────────────────────────────────────────────────── ─── Continued front page       (56) References Japanese Patent Laid-Open No. 2-116969 (JP, A)                 JP-A-4-222122 (JP, A)                 JP-A-7-239842 (JP, A)                 JP-A-3-273714 (JP, A)                 JP-A-6-204885 (JP, A)                 US Patent 5471412 (US, A)

Claims (10)

(57) [Claims] 1. A discrete cosine transform of six-stage operationQuick calculation
(DCTQuick calculation)Omitted the sixth stage internal multiplication operation ofSpeed
Based on arithmetic,For input data blockStrangeExchange day
A discrete cosine transform unit that obtains Connected to the discrete cosine transform unit,DC above
T speed calculation methodThe firstOne-round one-dimensional DCT transformSixth stage weight in
coefficientAnd the second one-dimensional DCT transformofDuring ~The firstSix-stage weighting factor
WhenScaling factor that is the product ofset(C (i,
j)) according to a given primitive quantizer matrix (Q
(I, j)), Q '(i, j) = Q (i, j) / C
Based on the modified quantizer matrix modified to (i, j)
You, Scaling conversion and quantization,Input data above
Quantization to obtain quantized DCT data corresponding to a subblock
A vessel, Cosine Transform and Quantum Used in Image Compressor with
Device. Wherein the upper Symbol conversion data is a two-dimensional transform data, the discrete cosine transform unit, the DCT Sokusan method first stage fifth stage including first round primary operations from the the input data block give the one-dimensional transform data I line the original transform operation, further including a Starring <br/> calculation of fifth stage from the first stage of the DCT Sokusan method from another direction with respect to the one-dimensional transform data perform a second round one-dimensional transform operation, cosine transform and quantization apparatus according to claim 1 to obtain the two-dimensional transform data. 3. A comprising a quantizer shadow picture compression device is connected with the discrete cosine transform unit and the discrete cosine transform unit, six stage computing the discrete cosine transform Sokusan method (DCT Sokusan method)
The inside of the first round one-dimensional DCT transform sixth stage weighting coefficient and scaling coefficient sets a multiplication product of the sixth stage weighting coefficients in the second round one-dimensional DCT transform (C (i, j)) A given primitive quantizer matrix (Q (i, j)) by
Modifying Q ′ (i, j) = Q (i, j) / C (i, j) to obtain a modified quantizer matrix, and omitting the sixth stage internal multiplication operation of the DCT speed calculation method described above. in, to build the discrete cosine transform unit, the first
Input data by the speed calculation method omitting the six-step internal multiplication operation
Obtaining a conversion data to blocks, configured the quantizer by the modified quantizer matrix
Built to perform scaling conversion and quantization, multiplication operations savings method of the input data block and the corresponding obtaining a quantized DCT data, the comprising at imaging compressor. 4. The upper Symbol conversion data is a two-dimensional transform data, the discrete cosine transform unit, a fifth step of Starring <br/> calculated from the first stage of the DCT Sokusan method for input data block the resulting one-dimensional transform data I including first round one-dimensional transform operation rows further from another direction from the first stage of the DCT Sokusan method with respect to the one-dimensional transform data of the fifth stage Starring <br /> have rows including the second round one-dimensional transform operation the calculation, is constructed so as to obtain the two-dimensional transform data, used for the scaling coefficient sets to compensate for the predetermined original quantizer matrix A method for reducing multiplication operations in an image compression apparatus according to claim 3, wherein 5. comprising an inverse discrete cosine transform unit shadow image compression device is further coupled to the inverse quantizer and inverse quantizer, the inverse discrete transform Sokusan method six stage operation (IDCT Sokusan method) < br /> round one-dimensional IDCT conversion in the first stage weighting factor and Purisukerin grayed coefficient set is the product of the first stage weighting coefficients in the second round one-dimensional IDCT transform in (P (i, j )) By a given primitive solution quantizer matrix (Q (i,
j)), and Q ′ (i, j) = Q (i, j) × P (i,
Fixed to j), and obtaining a corrected solution quantizer matrix, the inverse quantizer by the corrected solution quantizer matrix
And the solution quantization for the quantized DCT data
Performs a scaling transformation, and obtaining a solution quantized data was omitted first stage internal multiplication operations of the IDCT Sokusan method
Below, to build the inverse discrete cosine transform unit, the
From the solution quantized data by Sokusan method is omitted first stage internal multiplication operations, I corresponding to the quantized DCT data
The method of claim 3, further comprising the step of obtaining DCT data. Wherein said IDCT data is a two-dimensional inverse transformed data, the inverse discrete cosine transform unit, a sixth stage from the second stage of the IDCT Sokusan method with respect to the solution quantized data
Calculating to obtain a one-dimensional transform data I including first round one-dimensional transform operation the rows of further operation from the second stage of the IDCT Sokusan method of the sixth stage from another direction with respect to the one-dimensional transform data the have rows including the second round one-dimensional transform operation, IDCT
Built data to obtain, on kiss Kerin multiplication operations savings method of imaging compression apparatus according to claim 5, grayed coefficient set is used to compensate for the predetermined primitive solutions quantizer matrix. 7. inverse discrete cosine transform Sokusan method six stage computing first first round one-dimensional IDCT transformation (IDCT Sokusan method)
Step weighting coefficients and Purisukerin grayed coefficient set is the product of the first stage weighting coefficients in the second round one-dimensional IDCT transformation (P (i, j)) by a predetermined primitive solutions quantizer matrix (Q (I, j)), Q '(i, j) = Q (i,
j) × P (i, j) based on the modified solution quantizer matrix obtained by input, the input quantized discrete cosine transform
And inverse quantizer to obtain a solution quantized data through the scaling transformation data (DCT data) with a solution quantization, in conjunction with the inverse quantizer, a first of the IDCT Sokusan method
From above Kikai quantized data by Sokusan method is omitted stage internal multiplication operations, is used in the imaging compression and imaging decreased pressure comprising comprises the inverse discrete cosine transform unit obtained by IDCT data corresponding to the DCT data, the Inverse quantization and inverse cosine transform device. 8. The IDCT data is a two-dimensional inverse transformed data, the inverse discrete cosine transform unit, a sixth stage from the second stage of the IDCT Sokusan method with respect to the solution quantized data
Calculating to obtain a one-dimensional transform data I including first round one-dimensional transform operation the rows of further operation from the second stage of the IDCT Sokusan method of the sixth stage from another direction with respect to the one-dimensional transform data the performed including second round one-dimensional transform operation, IDC
It is built so that to obtain a T data, inverse quantization and inverse cosine transform apparatus according to claim 7 in which the scaling coefficient sets are used to compensate for the predetermined primitive solutions quantizer matrix. 9. The image compression apparatus comprises an inverse quantizer and an inverse discrete cosine transform unit connected to the inverse quantizer, and the inverse discrete cosine transform speed calculation method (IDCT speed calculation) of a six- step operation.
Method) First-stage weighting factor in the first-round one-dimensional IDCT transform
The number and first phase weighting coefficients in the second round one-dimensional IDCT transform
Purisukerin grayed coefficient set that is the product of the (P (i,
j)) is a given primitive solution quantizer matrix (Q (i,
j)), and Q ′ (i, j) = Q (i, j) × P (i,
j) to obtain a modified solution quantizer matrix , and constructing the inverse quantizer with the modified solution quantizer matrix.
Construction obtaining a solution quantized data through the scaling transformation by solution quantizing the quantized DCT data built, under omitting the first stage internal multiplication operations of the IDCT Sokusan method, the inverse discrete cosine transform unit to, above
From serial the solution <br/> quantized data by Sokusan method is omitted first stage internal multiplication operations, multiplication operations in the amount of coca obtaining a IDCT data corresponding to the DCT data, comprising a shadow image compression apparatus How to save money. 10. The IDCT data is a two-dimensional inverse transformed data, the inverse discrete cosine transform unit, a sixth stage from the second stage of the IDCT Sokusan method with respect to the solution quantized data
Calculating to obtain a one-dimensional transform data I including first round one-dimensional transform operation the rows of further operation from the second stage of the IDCT Sokusan method of the sixth stage from another direction with respect to the one-dimensional transform data performing a second round one-dimensional transform operation including, IDC
It is constructed so as to obtain a T data, multiplication operations savings method of imaging compression apparatus of claim 9 in which the scaling coefficient sets are used to compensate for the predetermined primitive solutions quantizer matrix.