TW398146B - Cosine transformation and quantification device, method for reducing multiplication operations of an image compressing device, inverse quantification and inverse cosine transformation device, and method for reducing multiplication operations of an image - Google Patents

Abstract

In order to reduce the multiplying operations in the image compressing device, the prescribed original quantifier matrix uses a multiplier transformation coefficient, which is derived from the sixth-level multiplication operation of a fast calculation method for discrete cosine transformation (DCT), as the compensation to obtain a corrected quantifier matrix. While the sixth-level internal multiplication operation of the DCT fast calculation method is omitted, a DCT unit is constructed to apply a DCT on an input data block following the DCT fast calculation method to obtain the transformation data from multiplier transformation. A quantifier is constructed to quantify the transformation data from the multiplier transformation according to the corrected quantifier matrix. In order to reduce the multiplying operations in the image decompressing device, the prescribed original de-quantifier matrix uses an advance multiplier transformation coefficient, which is derived from the first-level internal multiplication operation of a fast calculation method for inverse discrete cosine transformation (IDCT), as the compensation to obtain a corrected de-quantifier matrix. An inverse quantifier is constructed to de-quantify the transformation data from the multiplier transformation according to the corrected de-quantifier matrix. While the first-level internal multiplication operation of the IDCT fast calculation method is omitted, an IDCT unit is constructed to apply an IDCT on an output of the inverse quantifier following the IDCT fast calculation method.

Description

A7 printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs _; ____B? V. Description of the Invention (丨) This invention relates to image compression and decompression technology, and more particularly, it relates to an image that performs fewer multiplication operations Compression settings and image reduction devices. 5 Discrete cosine transform (DCT) is widely used in image compression calculations, such as MPEG 'for moving image coding and "Four (^) for still image coding. Similarly, inverse discrete cosine transform (IDCT) Corresponding inversion processing for decompressing compressed image data is used. Private IDCT and rDCT operations typically include several multiplication and addition operations. The multiplication operations through the DCT and IDCT processing are quite time-consuming and require quite complex hardware. In U.S. Patent No. 5,471,412, the 'applicant disclosed a six-stage DCT / IDCT rapid algorithm. For one-dimensional transformation, it only contains twelve multiplication vehicles, or for an 8x8 data environment. For the two-dimensional conversion, it only includes a total of 208 (2 × 8 × 13) multiplication operations. 15 The flow directions of the DCT and industrial DCT rapid algorithm are shown in the first and second figures respectively. In the image compression and / or image During the decompression, in order to obtain a higher processing speed, it is desirable to further reduce the number of multiplication operations. Therefore, `` the purpose of the present invention is to provide a method for using an image scaling device. Compared with the conventional technology, the cosine conversion and quantization device can perform fewer multiplication operations. Another object of the present invention is to provide a method for reducing the number of multiplication operations of an image compression device to obtain a higher number of multiplication operations. Method for processing speed. Another object of the present invention is to provide an inverse quantization and inverse cosine conversion device used in image compression and shadow forest decompression, which is applicable to the paper size on page 4 of the conventional technology, China® Country榡 Standard (CNS) 8 4 specifications (210 X 297 male (please read the precautions on the back before filling this page) Order: .. ¾ A7 B7 printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs l) In comparison, it can perform fewer multiplication operations. Another object of the present invention is to provide a method for reducing the number of multiplication operations of an image decompression device to obtain a higher processing speed. According to the present invention, The first feature is that the cosine 5 conversion and quantization device used in the image compression device includes a discrete cosine bar conversion unit and a quantizer. The discrete cosine conversion unit Performs a discrete cosine transform (DCT) on an input data block according to a six-level DCT rapid algorithm. The discrete cosine conversion unit omits the sixth-level internal multiplication of the DCT rapid algorithm to obtain a conversion that is multiplied. Data. The quantizer is connected to the discrete 10 cosine conversion unit, and quantizes the converted conversion data according to a modified quantizer matrix to obtain quantized DCT data corresponding to the input data block. The modified quantization The quantizer matrix is obtained by compensating a predetermined original quantizer matrix with a set of magnification transformations derived from a sixth-level internal multiplication operation of the DCT rapid algorithm. 15 According to a second feature of the present invention, a A method for reducing the number of multiplication operations of an image compression device is provided. The image compression device includes a discrete cosine conversion unit and a quantizer connected to the discrete cosine conversion unit. The method includes the steps of compensating a predetermined original quantizer matrix with a set of 20-fold conversion coefficients derived from a sixth-level internal multiplication operation of a six-stage discrete cosine transform (DCT) rapid algorithm. Correct the quantizer matrix; construct the discrete cosine transform unit without the sixth-level internal multiplication of the DCT rapid algorithm, and perform DCT on an input data block according to the DCT rapid algorithm, and obtain the magnification conversion Conversion data; and construct the quantizer according to the modified quantizer moment. Page 5 This paper size applies the Chinese National Standard (CNS) A4 specification (210X297 mm) (Please read the precautions on the back before filling this page )

Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs A7 B7 V. Invention Description (3)… to quantify the conversion data converted by the magnification to obtain the quantified DCT data corresponding to the input data block. According to a third feature of the present invention, an inverse quantization and inverse cosine conversion device used in image compression and image decompression includes an inverse quantizer 5 and an inverse discrete cosine conversion unit. The inverse quantizer board de-quantizes the input quantized discrete cosine transform (DCT) data according to a modified de-quantizer matrix to obtain the de-quantized data converted by the magnification. The modified de-quantizer matrix is obtained by A six-stage operation reverses the discrete cosine transform (I DCT) rapid algorithm. The pre-scale conversion coefficients derived from the first stage of the internal multiplication operation are compensated by 10-a predetermined original dequantizer matrix to be obtained. The inverse discrete cosine conversion unit is connected to the inverse quantizer and is operable to perform an industrial DCT on the multiplied converted dequantized data according to a six-stage operation IDCT rapid algorithm. The inverse discrete cosine conversion unit omits the first-level internal multiplication of the industrial DCT rapid algorithm to obtain industrial DCT data corresponding to the input quantized DCT data 15. According to a fourth feature of the present invention, a method for reducing the number of multiplication operations of an image decompression device is provided. The image decompression device includes an inverse quantizer and an inverse discrete cosine connected to the inverse quantizer. Conversion unit. The method includes the steps of compensating a predetermined original dequantizer matrix with a set of pre-scale conversion coefficients derived from a first-level internal multiplication operation of the inverse discrete cosine transform (IDCT) rapid algorithm of a sixteen-stage calculation. Obtaining a modified dequantizer matrix; constructing the inverse quantizer '俾 can dequantize the input quantized discrete cosine transform (DCT) data according to the modified dequantizer matrix to obtain the dequantized data after the magnification conversion; and the sixth The paper size of this page applies the Chinese National Standard (CNS) specifications (2 丨 0 · 〆297mm) (Please read the precautions on the back before filling this page) ^ Jv ----.--- Order — V mf ta §I ml m all

S 10 15 A7 B7 Under 'omitting the first-level internal multiplication operation of the IDCT rapid algorithm, construct a town inverse discrete cosine conversion unit to perform IDCT on the dequantized converted data to obtain the corresponding quantized DCT Information of information. Other features and advantages of the present invention will become clearer in the following detailed description of the preferred embodiment of the Central Standards Bureau of the Ministry of Economic Affairs of the People's Republic of China in conjunction with the accompanying drawings. The first figure is in the conventional art. The flow diagram of the six-stage DCT rapid calculation algorithm used in a one-dimensional transformation operation of a DCT_ _ unit; the second diagram is the sixth used in the conventional technique in one-dimensional reverse transformation operation of an IDCT unit. The flow diagram of the DCT rapid algorithm of the first-stage operator; the third diagram is a block diagram of a sound-decoding circuit of an image coding device incorporating a preferred embodiment of a cosine conversion and quantization device and an inverse quantization and inverse conversion device of the present invention; The fourth figure is a schematic circuit block diagram of the image decoding device of the embodiment combining the inverse quantization and inverse conversion device of the present invention; and the fifth figure depicts the operation of the cosine conversion and quantization device of the preferred embodiment of the present invention Flow chart; Quantitative conversion comparison of Shiling's private 11: Reversed quantization and inverse of the embodiment No. comparison meter (Please read the notes on the back before filling in this I.) πΐ!-Order-. Line 'paper carefully use China 苐 7 to buy X 297mm) 5. Description of invention (夂) 5 10 15 A7 B7 Printed by the Central Consumers Association of the Ministry of Economic Affairs Employee Cooperatives 20 Motion estimator · Discrete cosine conversion unit Reference frame buffer Inverse quantization and inverse cosine conversion device Inverse discrete cosine conversion unit 41 Adder 13 Variable-length browner Inverse quantization and inverse DCT device 70 The adder 6-inverts the DCT unit. The third figure is a schematic circuit diagram depicting an image coding device involving discrete cosine transform. -The motion estimator 1Q receives the source input frame and performs motion estimation related to a previous and / or a future input frame. If the input frame is a predictive encoding frame or a bi-directional predictive encoding frame (bi_directional predictije-coded), the motion estimator ι0 generates a motion direction 篁, the predictive encoding frame and the bidirectional prediction Both encoding frames are generally referred to as non-internal coding frames (n〇n__intra coded frames). Alternatively, for intra_coding frames, the input frames are directly奚 摈 to a cosine conversion and quantization device 20 ′ The device 20 includes a discrete cosine conversion (dct) unit 21 and a quantizer (Q) 22 connected to the DCT unit η. The ⑽ unit 21 enters this round. The frame performs discrete cosine transformation and provides the transformation data to the quantization H 22. The quantizer 22 is responsible for converting the transformation data from the cosine transformation and quantization device located at 10 21 30 40 4 2 12 50 60 80 62 quantizer subtractor 11 Inverse quantizer Variable length encoder reference frame buffer Inverse quantizer (Please read the precautions on the back before filling in this page) ----. Rj .- ^ ΛνΓ— \ / • ―———— Order --Chain __ ^ ---- i ..--- Page 8 Degree applicable to Chinese National Standard (CNS) 44 specifications (210x297 mm)

• t ^ n fm HI Printed by the Central Office Single Bureau MF Cooperative Agency 5 10 15 20 、 Explanation of the Invention (b) A7 B7 Amplitude of a set of limited levels in the coded picture. The non-processing is performed from the stored string conversion and quantization device 20, but only the reference image of the reference frame in which the prime value passes through-minus ^%, in the non-right meat group. 11 Subtracted from the pixel value of the current frame. If yes these: turn ::: turn = quantization device 2. Processing only—Inverse Han ^ ~ Inverted Lising and Inverse Cosine Conversion Device 40 includes a quantizer 运: Run: 丄 Early: The inverse quantizer 41 executes 21, wire, and seedling IDCT unit 42 executes the DCT unit器 12Μ 异。 12 different. These non-sensing-encoded maps are referenced by additions and inverse it. These reference frames can be re-constructed further from _code_, which is in ;; = reference frame buffer, or the == force = :: reference frame. In this case, the frame pirate 12 compensates the difference in pixel values of the predicted coded picture with the referenced picture frame of the predicted coded picture frame, and will newly construct the measured pictured reference picture frame buffer 30V. The operation system from the motion estimator 10 is caused to address the pixels of the reference frame within the reference frame buffer% to perform a subtraction operation, and the non-internal frame and the compensation operation can be performed.俾 The prediction coding reference frame can be re-framed ", ... -VLC 13 is connected to the output of the quantizer =, and is used to perform fman-resistant editing Entropy coding, 俾 this quantized conversion data to j .--.------ Γ.—rj --- (Please read the precautions on the back before filling this page) *-_1, «^ Ϋ i-line (page 9 of this paper size (CNS) secret (210 × 297 · ^ A7 B7 V. Description of the invention (Reflected to a variable length code that is continuously rotated out as a compressed image code. The motion estimator 10 shifts Go to the redundant counties between the frames of the input frames of these sources. The eliminator 22 removes the unimportant part of the image data, which is the naked eye The sense of law. The variable-length encoder 13 further compresses image data by applying consistent encoding technology, and can obtain compressed image data with compression ratios ranging from tens to hundreds. If these sources input If the frame is a static image without the relationship between the frames, like a photo, the image coding device of the first figure can be reduced to include only the cosine transformation and f makeup 4 2 * variable-length encoder & the motion Estimate n 1Q, which is used to provide the subtraction of the pixel value difference ^ η, which is used to provide the reference frame 12, the reference frame buffer 30 which is used to store the reference frame, and the inverse quantization and inversion The cosine conversion device 40 is omitted. &Quot; 'The above-mentioned general processing of the video signal encoding processing and encoding device is well known in the art, and it is also used in digital images such as MPEG * jpEG. Modification technology is widely used. Examples of encoding devices using cosine transform and quantization devices are disclosed in U.S. Patent Nos. 5'74 8,240, 5,321,725, and 4,245,24 8. The fourth picture series Draw a schematic circuit block diagram of an image decoding device corresponding to the image encoding device of the third figure. The image decoding device includes a variable length decoder (VLD) SO, an inverse quantization and inverse cosine conversion device, and a parameter. The frame buffer 70 and an adder 80. The device so includes an inverse quantizer, (DeQ) 61, and a one-turn discrete cosine transform (IDCT) unit M connected to the inverse quantizer 61. This variable-length scorpion eliminator 〇 Processing jI—i xi iio —--- (Please read the precautions on the back before filling this page)

Printed by the Consumers' Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs J 1Ε—Order-One I —- I________ Page 10 This paper size is applicable to China National Standards (CNS) 々4 specifications (210X297 mm) 5. Description of the invention (A7 B7 5 10 15 Printed by the Central Mongolian Bureau of the Ministry of Economic Affairs f Cooperatives 20 These compressed image codes are variable-length codes, which can be used as data. The inverse quantizer 61 calculates the quantized DCT data to obtain solutions. Quantitative DCT data. The implementation of IDeT for the quantized DCT data of Unit 2 and 2 Quantizer 6! Is based on the case that the industrial and commercial data are used in the inline side frame, and in the case that is not included in the inline Mart frame. It is for the difference. = The work of the reference frame buffer H 7Q is the same as the reference frame buffer 3Q in the image code device. The addition code is from the reference frame buffer that is not the internal encoding frame. 7. Reference frame of the device ^ The data of the plate T. The internal frame from the IDCT unit s2 and the internal frame of the code are used to create the reconstructed frame ^ The internal frame of the restructured table and the frame The brown frame of the predictive tank is also stored in the reference frame buffer 70. Reference frame. The reference frame = 7. It has a motion vector turn-in, and the motion vector input is two f dream test frame buffer i 7. The appropriate pixel addressing of the ship reference frame is being rebuilt. The compensation operation is handled when the frame is edited. If the image code for the money return only contains the secret image without the frame and the eye-to-eye relationship, like the photo, the image decoding device in the fourth image can be reduced and included. ㈣Long material correction and inverse conversion device ... It is used to reconstruct the non-internal coded picture plagiarism 80 and the reference frame buffer 70. The general description of the decoding processing and decoding device of the video signal described above The concept is also well-known in the conventional technology, and it is widely used in digital image compression technology such as fresh and key reading. _____ Page U The paper size is quiet 11¾ standard (c is called $ 4 秘 (21Gx297 公(Please read the precautions on the back before filling out this page} • 01. Order-Line '• I-II-1- * .I 1 Printed by the Consumers' Cooperative of the Central Standards Bureau of the Ministry of Economic Affairs A7 B7 5. Description of the Invention (1)-In a cosine conversion and quantization device, the quantity The inversion operation is often followed by the discrete cosine conversion operation. Furthermore, in an inverse quantization and inverse cosine conversion device, the inverse discrete cosine conversion operation is often followed by the inverse quantization operation. 5 US Patent by the Applicant The DCT / IDCT device disclosed in Case No. 5,471,412 performs DCT on an 8x8 input data block based on the six-level operation DCT rapid algorithm shown in the first figure, and is based on the display shown in the second figure The six-stage operation IDCT rapid algorithm performs DCT on an 8x8 input conversion data block. 10 As shown in the first figure, the six-level operation system of the DCT rapid algorithm includes a first-level operation including four butterfly operations, and a second-level operation including two multiplication operations after addition. Operations, a third-level operation with four butterfly operations, a fourth-level operation with three multiplication operations after addition, a fifth-level operation with four butterfly-type operations, and 15 one with eight Sixth level of internal multiplication. The two-dimensional DCT includes consecutive first and second rounds of one-dimensional DCT operations. The second round of one-dimensional DCT operation is performed in the other direction on the one-dimensional transformation data from the first round of one-dimensional DCT operation, and two-dimensional transformation data corresponding to the input data block is obtained. As shown in the second figure, the six-level 20 operation system of the IDCT rapid algorithm includes a first-level operation including eight internal multiplication operations, a second-level operation including four butterfly-type operations, A third-level operation including three multiplication and subtraction operations, a fourth-level operation including four butterfly operations, a fifth-level operation including two multiplication-subtraction operations, and a four-level operation Sixth level of butterfly operation. The two-dimensional DCT includes page 12. This paper size is applicable to the Chinese National Standard (CNS) today 4 specifications (210X297 mm) (Please read the precautions on the back before filling this page)-Line ί — A7 B7 5 10 15 Economy Printed by the Zhengong Consumer Cooperative of the Ministry of Health, the Central Bureau of Justice, 20.7, V. Description of Invention (\ 〇

Lian: The first and second rounds of dimensional IDCT operations. The second round-dimensional IDCT is performed in another direction on the one-dimensional reverse transformation data from the first round of one-dimensional DCT operation, and obtains the two-dimensional reverse transformation data corresponding to the input transformation data block. In the quantizer of the cosine transform and quantization device, the quantization operation uses a predetermined quantizer matrix suitable for different image attributes to perform an 8 × 8 square. That is, 'the predetermined quantizer matrix may be different due to the difference between brightness and color user f, due to internal coding and non-internal coding frames, and so on. One image sequence to another image sequence. The difference and special 4. A basic equation of the refining operation is: M ^ 〇) = £ (1,3) / 0 (1,3) where L j f (1 ^ j) represents DCT data,

Qii ': n represents the predetermined quantizer matrix M (i / j) represents the quantization conversion asset. The inverse quantization and inverse cosine conversion are installed in the inverse quantification _ with-dare 'box basis to perform. "Basic Equations for Chemical Operations ^ To a 8x8 f, j) = M (i, j) * Q (i / j} where = 0,1, —, f (i / j) represents the dequantized conversion data , Q (i ']) represents the predetermined dequantizer matrix, and Mk, j) represents the quantized conversion data. Please refer to the third circle, “different from conventional technology” again. The present invention is more suitable for the national standard of this paper. (Please read the precautions on the back before filling in this page}

And on page 13, the Consumer Consumption of the Central Standards Bureau of the Ministry of Economic Affairs prints A7 B7 ______ V. Description of the Invention () \…… The cosine conversion and quantization device 20 of the preferred embodiment is characterized by the DCT. Unit 21. 在When the sixth-level internal multiplication operation of the DCT rapid algorithm is omitted, DCT is performed on the input data block according to the DCT rapid algorithm to obtain the conversion data after the magnification conversion. Based on this, the quantizer 22 is constructed. It can quantize the converted data converted by the magnification according to a modified quantizer matrix, and can obtain the quantized DCT data corresponding to the input data block without any performance degradation. The modified quantizer matrix is obtained by compensating the predetermined original quantizer matrix with a set of magnification conversion # numbers derived from a sixth-level internal multiplication operation of the DCT rapid algorithm. The overall processing speed of the cosine transform and the quantization device 20 is therefore greatly improved. By omitting the sixth-level internal multiplication operation of the DCT rapid algorithm, the number of multiplication operations performed by the DCT unit 21 can be greatly reduced. In particular, for one-dimensional DCT conversion, the number of 'multiplications' is 8x; the total number of input data blocks is reduced from thirteen to five, that is, for one-dimensional DCT conversion of magnification, only Keep about 38% of the multiplication operations. As for the two-dimensional DCT conversion, to complete the conversion of the 8x8 input data block, the two-dimensional DCT only needs a total of 80 (2xSx5) multiplication operations. The conversion data from the DCT unit 21 of the preferred embodiment of the present invention is two-dimensional conversion data. The fifth figure is a flowchart for describing the operation of the cosine to 20 conversion and quantization device 20. First, the DCT unit 21 executes the first round of one-dimensional conversion operations on the input data blocks including the first to fifth stages of the DCT rapid calculation algorithms shown in the first figure, and obtains the first converted one. Dimension conversion data. Thereafter, the DCT unit 21 executes the one-dimensional conversion data converted by the magnification in the other direction to include the _; ____ Page 14 Chinese country 棹 ^ " ~~-: — (Please read the precautions on the back before Fill out this page) --------- Order --- k ·, V. Description of invention (\ 乙

5 10 15 Printed by the Consumer Cooperatives of the Central Bureau of Standards, Ministry of Economic Affairs. 20 Nose method. The first to fifth level of the conversion of the magnification. The second round of the conversion is to obtain the two-dimensional conversion data after the conversion of the magnification. Then, according to the modified quantizer matrix, the dimensional spinner ^ 2 changes the two-dimensionally transformed data scene data block DCT data block M to the two-dimensional quantization corresponding to the input data block. The benefit matrix Q, (i'j) is obtained from Q according to the following equation, (l〇) = Q (i, j) / c (i, j) Ganshi Q (Tor '") represents a predetermined original quantization Filter matrix, and C (i, j) ratio conversion coefficients. As shown in the sixth figure, 'The set of multiples of one, four, and C (i, j) is the first step in the first round of one-dimensional DCT conversion and the DCT rapid algorithm in one-dimensional DCR conversion. Product of six weighting coefficients. The Lihua formula may be combined with other items, such as the person who takes the money conversion into consideration, and the merge-rate converter to the quantized moment beer (for example, the quantizer ratio conversion in MpEG) (guantizer scale)). If there are combined items, the group of conversion coefficients of the group need to be assigned to these items before the quantization operation is performed. According to this, in order to reduce the multiplication operation of an image compression device including a DCT unit 21 and a quantizer 22 connected to the DCT unit 21, a pre-trained original chemizer matrix is dispersed in a group from a six-level operation. The multiplication factor derived from the sixth-level internal multiplication operation of the cosine (DCT) rapid algorithm is used for compensation, and a modified quantizer matrix is obtained. After that, the 15-page paper size was exempted from the application of the Chinese national standard (CNS > A4 size (2 丨 0X297 public holiday) (Please read the precautions on the back before filling this page). Order the staff of the Central Bureau of Standards of the Ministry of Economic Affairs Cooperative printed A7 ___ B7 V. Description of the invention € θ)-: Under the sixth-level internal multiplication of the DCT rapid algorithm, the DCT unit 2m is constructed to perform DCT on an input data block according to the DCT rapid algorithm,俾 Get the conversion data converted by the magnification. The quantizer 22 is structured to quantize the conversion data converted by the magnification according to the modified quantizer matrix to obtain quantized DCT data corresponding to the input data block. If so, the quantization operation incurs a small increase in operation time. However, the DCT operation has a dramatic reduction in operation time. Therefore, when the cosine transform and quantization device 20 is used during image compression, the overall performance is improved. 10 Please refer to the third and fourth figures again. The inverse quantization and inverse cosine conversion device 40'60 of the preferred embodiment of the present invention is characterized in that the inverse eliminators 41 and 61 are configured according to— The dequantization matrix is modified to negativeize the DCT of the input. It is expected that the solution will be straightened to convert the quantization data to the ratio. It is expected that the positive dequantization matrix is obtained by using a set of first-level I5 algorithms from the IDCT rapid algorithm. The pre-magnification conversion coefficient derived from the internal multiplication operation is obtained by supplementing a predetermined original dequantization matrix. After that, the first-level internal multiplication of the DCT rapid algorithm is omitted. 'The IDCT units 42, 62 perform IDCT based on the six-level arithmetic IDCT. The rapid algorithm performs IDCT on the quantized data.' IDCT data corresponding to the input quantized DCT data 20 is obtained without performance degradation. The reversal of the chords and drags ^ ^ ^ The overall speed i. Therefore greatly improved. By omitting the first-level internal multiplication operation of the DCT rapid algorithm, the number of multiplication operations performed by the IDGT unit 4: 2,62 can be greatly reduced. Specially, in terms of the one-dimensional DCT conversion, the total number of multiplication operations is on page 16. This paper size applies the Chinese National Standard (CNS) 8-4 specification (210X297 mm) (may read the precautions on the back before filling this page) } ---- «—f ^ — ~ V ml tm- ·, ι-β -line. Five A7 B7 5 10 15 Printed by the Consumers' Cooperative of the China Standards Bureau of the Ministry of Economic Affairs 20, Invention Description (\ M *) One 8-parent 8-input conversion data block is reduced from thirteen to five, that is, for the rate conversion-dimensional IDCT, only about 38% of the multiplication operation. As for the two-dimensional IDCT conversion, the 8 young input must be completed The conversion of the data block to convert the two-dimensional DCT only requires a total of S0 (2X8X5) multiplication operations. Second, the IDCT data from the 1DCT unit 42/62 of the preferred embodiment of the present invention is converted data for the two-dimensional DCT. The seventh diagram is a flowchart depicting the operation of the inversion conversion and the inverse cosine conversion device 40, S0 of the third and fourth diagrams. First, 'the inverse quantizer 41, 61 dequantizes the II matrix according to the correction to turn the round into Dequantize the quantized DCT data block, get the magnification conversion and dequantize the data ^ and then 'The IDC The T units a, a perform the first-round one-dimensional reversal conversion operation of the second-to-sixth-level operations of the multi-level conversion of the DCT de-quantized data, and the one-dimensional reversal conversion of the multiplication ratio can be obtained. ^ Data. The IDCT unit 42, a further performs one-dimensional reversal conversion on the magnification conversion data in another direction. The magnification conversion including the first to sixth stages of the IDCT rapid algorithm is performed. I know the IDCT data corresponding to the input quantized DCT data block. The modified dequantizer matrix Q is obtained according to the following equation: Q (i / j) == Q (i, j). * P (i, j) where 'Q (i, j) represents the predetermined original dequantizer matrix' and p (1,: ί) represents the set of pre-scale conversion coefficients. As shown in the eighth figure ' The set of pre-magnification conversion coefficients p (iij) is the product of the first-stage weighting number of the IDCT rapid differentiation method in a first round / dimensional IDCT conversion and a second round one-dimensional IDCT conversion. The paper size of this page applies to the Chinese National Standards (CNS) A4 rules. ¥. (21〇χ297 mm) (Please read first and back (Please fill out this page again)

Five A7 B7 5 10 15 Ministry of Economic Affairs A Associate Bureau. Staff W cooperation. Printed by the agency 20 Exam ::: and ,: Resolve and resolve to this quantization matrix (for example, / MD parallel port conversion (cm as' The quantizer scale in SG) The sheep conversion pre-scale conversion coefficients are used in the dequantization operation: line:!: Words' the group to these items. Executive Zhigang also needs to be assigned accordingly, In order to reduce the wall including a reversal to read reversal quantizer 4lfSl, the production of 41, 61 and-multiplication operation of the continuous operation device,-early Wu 42'62 video decompression ~ six-level operator two call dequantizer The matrix is derived from one or two sets of pre-multiplying conversion coefficients as the second and second-level internal multiplication operations, and then the inverse _ positive solution quantizer matrix. The l61 system can be constructed according to this correction. Data: In: : DCT data is dequantized to get the magnification conversion solution. I4 is the first-level internal multiplication operation of the IDCT rapid algorithm ^ 'The IDCT unit 42 "62 is constructed to quickly perform the quantization data execution based on the work_ 1DCT, you can get IDCT data corresponding to the rotation of 1D DCT data.疋 The inverse quantization operation incurs a small increase in operation time. However, the DCT operation in this process has a dramatic reduction in operation time. Because the inverse quantization and inverse cosine conversion device 4Q " 6. In the image The overall performance is improved when used during dust removal and Λ. Although the present invention has been described in conjunction with the preferred embodiments, it should be understood that the present invention is not limited to the disclosed embodiments and is intended to cover Various changes and equivalent embodiments are included in the spirit and scope of the present invention.

Please read the precautions on the back first, fill in this book '} tr -----

Claims (1)

Printed by A8 'B8 C8 D8 of the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs 6. Application for Patent Scope 1. A cosine conversion and quantization device used in an image compression device, including: a discrete cosine conversion based on a six-level operation ( DCT) rapid algorithm to perform DCT discrete cosine transform unit 5 on an input data block, the discrete cosine transform unit omitting the sixth-level 'internal multiplication operation of the DCT rapid algorithm to obtain conversion data converted by the magnification; and A quantizer is connected to the discrete cosine conversion unit, and quantizes the converted conversion data according to a modified quantizer matrix to obtain quantized DCT 10 data corresponding to the input data block. The modified quantizer matrix is obtained by compensating a predetermined original quantizer matrix with a set of magnification conversion coefficients derived from the sixth-level internal multiplication operation of the DCT rapid algorithm. 2. The cosine conversion and quantization device described in item 1 of the scope of patent application, wherein the conversion data converted by the magnification is two-dimensional conversion data, and the discrete cosine conversion unit executes the input data block including the DCT rapid algorithm's first to fifth level operation of rate conversion The first round of one-dimensional conversion operation obtains the rate conversion one-dimensional conversion data, and further executes the rate conversion one-dimensional conversion data in another direction including the DCT quickly The first to fifth stages of the algorithm's magnification conversion are performed in the second round of the 20-dimensional one-dimensional conversion operation to obtain the magnified conversion data. 3. A method for reducing multiplication of an image compression device, the image compression device comprising a discrete cosine conversion unit and a quantizer connected to the discrete cosine conversion unit, the method comprising the following steps: Six-stage calculation of discrete cosine transform (DCT) rapid calculation page]. 9 pages (please read the precautions on the back before filling this page) I The paper size of this edition applies the Chinese National Standard (CNS) A4 specification (210X297 mm) 5 1 〇15 The scope of the patent application A8 B8 C3 D8 Printed by the China National Standards Bureau Employees Cooperative of the Ministry of Economic Affairs. 20 The multiplication factor derived from the six-level internal multiplication operation is used to calculate the quantizer matrix. A modified quantizer is obtained. Quick calculation (sixth level) multiplication to build the discrete hiding conversion unit to perform DCT on the second round of ^ data blocks according to the dgt fast algorithm. The quantizer matrix is modified to quantize the converted data of the magnification conversion to obtain and quantify the DCT data. (4) The method described in item 4 of item 3 of the input data block, wherein the conversion conversion of the magnification conversion is two-dimensional conversion data, and the discrete cosine conversion unit is constructed to the input data block. Execution 2 contains the rate conversion from the _th to the first operation of the speed-reduction algorithm of the town-the first round-the conversion operation to obtain the conversion of the conversion-_ change the data,-step the conversion of the conversion in another direction-the implementation of the dimensional conversion data ^ 3 The rate conversion of the first to fifth level operations of the DCT rapid algorithm. The second round of the i conversion operation to obtain the converted data of the rate conversion, " The group used to quantify the H material towel in the compensation_Dingzhi County The magnification conversion coefficient is a product of the sixth-stage weighting system ^ of the DCT rapid algorithm in the -first round_dimensional DeT conversion and the second round one-dimensional DCT conversion. 5. According to the method described in item 3 of the application for the special township, the shadow reduction device further includes-an inverse quantization reversal-reversing cosine conversion unit of the quantizer, wherein the method further includes the following steps : Page 20 The size of this paper mark applies to the Chinese National Standard (CNS) Α, 4 (ZlOxm mm) (read the notes on the back of the poem before filling in this page) A 8 B8 C8 D8 'The scope of patent is 5 ίο 15 Printed by the Central Government Bureau of the Ministry of Economic Affairs of the Employees' Cooperatives 20 Algorithms and / or Levels Differential Inverse Discrete Cosine Transform (IDCT) Quick Compensation-Pre-multiplier derived from internal multiplication The conversion coefficient is used to transform the initial dequantizer matrix, and 俾 can be obtained-corrected solution amount. The f i1 is reversed to the I quantizer. The modified dequantizer matrix can be used to dequantize the CT breeding material to obtain the quantized conversion dequantization data. Based on the first-level internal multiplication operation of the 1DCT rapid algorithm, construct the inversion hedge method to convert the magnification conversion unit to perform the DCT conversion based on the rapid calculation of the DCT calculation, and obtain the data corresponding to the DCTt. The function information. The method described in item 5 of the scope of the second patent application, wherein the IDCT data is a two-dimensional inverse conversion data, and a θ, a discrete discrete cosine conversion unit is constructed to execute the conversion data of the magnification conversion. Contains the second to sixth stages of the IDC: T rapid algorithm. The first round of one-dimensional inverse conversion operation is performed to obtain the magnification conversion, the dimensional reverse conversion data, and the magnification conversion is further performed in another direction. The dimensional inverse conversion data performs the scaling conversion including the first to sixth operations of the IDCT rapid algorithm. The second round of one-dimensional inverse conversion operation obtains the IDCT data, and is used to compensate the predetermined original dequantizer matrix. This set of pre-magnification conversion coefficients is the product of the first-stage weighting coefficients of the idcT rapid algorithm in a first round of one-dimensional IDCT conversion and a first round of 1DCT conversion. Page 21 (CNS) Eight Fortune Samples (2i〇X297 Gong Ding Ding (Please read the precautions of Xixi before filling out this page) Printed by the Consumer Cooperatives of the Central Standards Bureau of the Ministry of Economic Affairs A8 B8 C8 D8 6. Scope of patent application 7. A reverse quantization and reverse cosine conversion device used in image compression and image decompression, including: A quantizer matrix to dequantize the input quantized discrete cosine transform (DCT) data to obtain the dequantized 5 quantized data inverse quantizer. The modified dequantizer matrix is a set of inverse discrete A pre-scale conversion coefficient derived from the first-level internal multiplication operation of the cosine transform (IDCT) rapid algorithm is obtained by compensating a predetermined original dequantizer matrix; and a connected to the inverse quantizer is operable to operate according to the The six-stage 10-operation IDCT rapid algorithm performs an IDCT inverse discrete cosine conversion unit on the dequantized data that has been converted by the magnification; the inverse discrete cosine conversion unit omits the first-level internal multiplication operation of the IDCT rapid algorithm to obtain a value corresponding to the Enter the IDCT data of the quantified DCT data. 8. The device for inverse quantization and inverse cosine conversion 15 described in item 7 of the scope of the patent application, wherein the IDCT data is two-dimensional inverse conversion data, and the inverse discrete cosine conversion unit executes the quantization data for the ratio conversion. The first to the first round of one-dimensional reversal conversion operation to obtain the data of the one-dimensional reversal conversion of the magnification conversion, and further to the one-dimensional inverse 20-reversion conversion of the magnification in the other direction. The data performs the magnification conversion including the second to sixth level operations of the IDCT rapid algorithm. The second round of one-dimensional inverse conversion operation is performed to obtain the IDCT data. The set of presets used in compensating the predetermined original dequantizer matrix The magnification conversion factor is for the first round of one-dimensional IDCT conversion and one page 22. This paper size applies the Chinese National Standard (CNS) A4 specification (210X297 mm) (Please read the precautions on the back before filling this page) Bu-Q. Order 5 10 15 Scope of patent application A8 B8 C8 D8 Member of the Central Bureau of Standards of the Ministry of Economic Affairs X Consumer Cooperative Co., Ltd. Printed 20 rounds of IDCT transformation in the maintenance of IDCT The product of the first stage speed weighting coefficient "Tony different Law. A method for reducing the multiplication operation of an image decompression device. The image decompression device includes an inverse quantizer and an inverse discrete cosine conversion unit connected to the inverse quantizer. The method includes the following steps: Six-stage operation reverses the discrete cosine transform (IDCT) rapid decision algorithm. The first-stage internal multiplication operation derives a pre-scale conversion system. A predetermined original dequantizer matrix is obtained to obtain a modified dequantizer matrix. According to the modified dequantizer matrix, the round quantized discrete cosine transform (DCT) data can be dequantized to obtain the dequantized data after the magnification conversion; and the first-level internal multiplication of the IDCT rapid algorithm is briefly described. Under operation, β constructs the inverse discrete cosine transform unit to perform IDCT on the dequantized data that has undergone the magnification conversion to obtain IDCT data corresponding to the input quantized DCT data. 10. The method described in item 9 of the scope of patent application 'wherein the]: the DCT data is a two-dimensional inverse conversion data, and the inverse discrete cosine conversion unit is constructed to perform the dequantization data on the magnification conversion to include the] IDCT's rapid algorithm's second to sixth level of magnification conversion. The first round of reverse conversion operation to obtain the magnification conversion one-dimensional reverse conversion data, and further performing the magnification conversion one-dimensional reverse conversion data in another direction. The second-to-sixth-level operation of the IDCT's rapid algorithm. The second round of one-dimensional inversion. Page 23 This paper size applies the Chinese National Standard (CNS) A4 specification (210X297 mm). Please fill in this page) i --- if # 'A8 B8 C8 D8 VI. Apply for patent range conversion operation to obtain the IDCT data, and use this set of pre-magnification conversion to compensate for the predetermined original dequantizer matrix The coefficient is the product of the first 5-level weighting coefficients of the IDCT rapid algorithm in a first round of one-dimensional IDCT transformation and a second round of one-dimensional IDCT transformation. (Please read the notes on the back before filling out this page) Printed by the Consumer Cooperatives of the Central Bureau of Standards of the Ministry of Economic Affairs Page 24 This paper size is applicable-National Code of Practice (CNS) A · 4 (210 X 297 mm)