TWI247246B - Scaling device and method for scaling a digital picture - Google Patents

Abstract

A scaling device and method for scaling a digital picture in only one pass requires only a small amount of buffer memory which can be used in real time. The method includes inputting the digital picture, generating a block of the digital picture from the digital picture, creating a first weighting matrix and a second weighting matrix, and multiplying the block of the digital picture by the first weighting matrix and the second weighting matrix.

Description

Japanese revision (more) is replacing the page number: 093111793 July 2014 I2 revision, invention description: [Technical field of the invention] The present invention provides a scaling device for zooming a digital picture and related methods, especially A scaling device that scales a digital picture and associated method that utilizes only one conversion to handle horizontal scaling and vertical scaling, and that only a small portion of the buffer memory is needed to achieve a scaled digital picture. ° [Prior Art] The traditional digital image scaling technique divides the two-dimensional scaling processing into two-dimensional scaling processing. In other words, the two-dimensional scaling processing first performs horizontal (width) scaling processing and then vertical (height) scaling processing. Theoretically, the equation is subsampled to obtain a new sampling point to perform a -dimensional scaling 40= (1) is the translation of the original pixel value by the cable U at the index n, h ( η) Numerical value. Furthermore, (4)·β insert (interP°lati〇n funCtion) = two total use _ original pixels, (9) weighting). The weight of the series is the index η (tap one, the two scaling of the traditional scaling method because of the need to wait for the horizontal scaling process completely traditional scaling technology has two shortcomings, the principle is not applicable to the instant application level, 1247246] -. Case No.: 093111793 The vertical scaling process can be performed after the correction is completed on July 12, 1994, or the horizontal scaling process must be performed after the vertical scaling process is completely finished. In addition, due to the two scaling processes, the traditional scaling technique Buffer memory is required to store horizontally scaled images and buffer memory is required to provide any interpolation filter of any length to achieve the desired zoom quality. Figure 1 is a conceptual diagram of a conventional magnification process, in which WoH And Hoid is the width and height of the original digital image, and Η _ is the width and height of the scaled image. The buffer memory system is used to store the horizontally scaled image, and the buffer size is (boundary XH〇id) Bits. Traditional scaling techniques have a number of disadvantages, especially magnifying the picture. Suppose the width and height of the original image being input is high. If you zoom in twice, the required buffer size is (Wnew X H_) = 2 X (W〇id X Hold). In addition, the delay time of traditional scaling technology cannot be applied to some applications, so the demand for large memory capacity It is not suitable for hardware design such as integrated circuit (1C), and the long data delay time is not applicable to the immediate application level. In order to solve the above problems, the general approach is to reduce the quality of vertical scaling. In order to reduce the buffer memory requirements, it means to reduce the data delay time. For example, the vertical scaling process uses only one or two filters, so only one buffer of two rows is needed. The required buffer will reduce the quality of the zoom, and therefore there is a need for a device and method that can be applied to an instant application, with high quality of zoom and low memory requirements. SUMMARY OF THE INVENTION The present invention provides a zooming device that can zoom a digital picture and Related methods to solve the above problems. 1247246 ; : . July 12, 1994 amendment. 94. „7. 1.2 — J , ... a case number: 093111793 The invention discloses a scaling device for zooming a digital picture, which comprises an original buffer for storing the digital picture, a processing unit for generating a Jth; a weighting matrix and a second weighting matrix, and a middle a buffer for storing intermediate data obtained by multiplying the first weighting matrix by the block of the digital image, and a target buffer for storing the second weighting matrix multiplied by the middle of the intermediate buffer The obtained output tribute. The method for scaling a digital image includes rotating the digital image, generating a block of the digital image according to the digital image, generating a first weighting matrix and a second weighting matrix, and The block of the digital picture is multiplied by the first weighting matrix and the second weighting matrix. [Embodiment] FIG. 2 is a conceptual diagram of a zooming technique 200 of the present invention, which is an enlargement process. In the zooming technique, the width (Wcld) and the height αι〇 of the digital picture 210 are first captured, and the digital picture 21〇 is vertically and horizontally scaled by the interpolation filters of length m and n, respectively, to continuously generate A plurality of blocks 220 each having a size of m χ 2' and each block 220 is multiplied by two weighting matrices to obtain a scaled picture 230 of l X η (4), and the reduction process utilizes the same concept. The invention reduces the complexity of the silk enthalpy, and the invention has two weighting matrices for storing the weighting coefficients of the waver series, and the weighting coefficients are the coefficients of the precision calculation, which can be appropriately adjusted and read and combined. ", and in the illusion of the equation, the ^ of the seven is divided into several sections, each - with the system represents all the sampling of the extent of this segment = other 'the invention of the secret S to provide the Mm series, therefore, With the "high precision required of the present invention, the range of t must be precisely divided into sections, and the number of filter stages required will increase; conversely, if the system I2fZ24f case number: 093111793 p The condition of the = system required for the amendment on July 12, 1994, then the range of t only needs to be divided into fewer chopping waves, and the number of series will be reduced. Segment,: ί" 流程图 weighting matrix flow chart. Suppose that the range of 'divided into h areas is L, the scaling factor is S, and ^ b is the weighting coefficient, ^3, enter 3, 3, 1), 11 and ^ Step 33 () in the = standard AdjustEn = 〇 (disable), and in step (four) if s is less than wide then iTml- Sa = 15 5 S " S ° ^ AdjustEn = 1 (enable) * If in step 340 If s is smaller than 卜, adjust the scaling factor (4). Otherwise, Sa=Qin establishes the weighting matrix WeightMat in step 360 as follows:

Hhj)

WeightMat(i,j)=酽(,·,··) where i ranges from 0 to (Η-1), and j ranges from 〇 to (L—υ, w(i, 』·) is at ( 1, j) is the weighting of the normalization first, and W(i, j) is the normalization coefficient, as follows: W(i,j) = 2_, /) The weighting w(i,j) of the first normalization is as follows:

Hhj) = h P a j + Η sa where h(s) is an interpolation filter, p=L/2~l, and H represents the number of segments of the division t. Sa is the adjustment scaling factor. FIG. 4 is a flowchart of the zooming technique of the present invention. In step 410, the digital picture 21〇 is input and stored in an original buffer, and the size of the original buffer is at least a digital picture 1247246 wide \ 嗔 '丨 2 ' ... case number: 093111793 The size of the height and width of the 210 was corrected on July 12, 1994. Step 420, determining the horizontal and vertical scaling factors according to the final image size required by the user, and the user inputs the required image size, and the scaling factor is respectively divided by the width and height of the digital image 210 by the width of the rotated image 230. And the height is obtained, and two weighting matrices are established according to the scaling factor and the flowchart of FIG. In steps 430 to 450, new sampling points are scanned according to the scaling factor, and the scanning of the new sampling points includes the following steps: determining the pixel index required for the interpolation filter in step 430, and adjusting the indexes according to the boundary conditions; The block 220 of the digital picture 210 is stored in the block buffer according to the indexes; the step 450 performs horizontal and vertical scaling according to the matrix multiplication result; the step 460 inputs the scaling result of the step 450 to a target buffer; In step 470, steps 430 through 460 are iteratively repeated until all new sampling points have been generated. If all new sampling points have been generated, the process ends; if some new sampling points have not been generated, then jump to step 430 and repeat steps 430 through 450. In a preferred embodiment of the invention, t is divided into 16 segments (H = 16) and 4 filter stages (L = 4) are used to illustrate the preferred embodiment of the invention, and the number of segments and The number of filter stages varies from design to design and does not limit the application of the invention. The flow chart detailed in the present invention is divided into three parts, namely, the initialization of FIG. 5, the vertical scaling of FIG. 6, the horizontal scaling of FIG. 9, and FIG. Referring to FIG. 5, the digital picture 210 is first input into a raw buffer SRC in step 510. The size of the digital picture 210 is W-XH〇ld, and the size of the original buffer SRC is sufficient to accommodate the digital picture 210. The size is; in step 520, the user inputs the desired new image sizes Wnew and Hnew, and the horizontal and vertical scaling coefficients are respectively obtained by the following equation: _ Wold Hold

Sh^W a Sv~ Η ff new and new where Wold and Ilid are the width and height of the digital picture 210, respectively, Wnew and Hnew 1247246 .Μ·.7· Γ2 Case number: 093111793 July, July 2nd, respectively The width and height of the picture 230 after scaling. When the scaling factor has been determined, the horizontal matrix WeightMat2 and the vertical matrix WeightMatl are generated according to the flowchart of Fig. 3 and the equations previously described. In this embodiment, since H=16 and L=4, the sizes of WeightMat1 and WeightMat2 are both 16x4. Note that the size of the weighting matrix differs according to the required resolution, and the values are Η and L. To make adjustments. A large Η value indicates that the range of t is divided into more segments, meaning that the resolution of the scaling is higher, and the number of filter stages corresponding to a larger L value is larger, meaning that there are more environmental pixels in the reference. When the weighting matrix WeightMatl and WeightMat2 are established, the setting of the scaling factor flag AdjustEn is not set to 1 (enable), and the setting is adjusted according to the application level. In the final step 530 of the initialization process, a new vertical index y2 is set to 0 before the vertical scaling is performed, and the flowchart of FIG. 5 is connected to the connection A in FIG. 6 to continue to represent the entire process. . Please refer to FIG. 6 , which is a flow chart of the vertical scanning process. The flow chart is continued by connecting A to FIG. In step 610, the relationship between the old vertical yl of the original image and the new vertical index y2 of the scaled image is defined by the following equation: yl = Sv * y2 y = floor(yl) wi_v=round[(yl- y)xH] where sv is the vertical scaling factor, y2 is the new vertical system of the scaled image is the old vertical index corresponding to the input original image of y2: 〇 71 y is the answer part of yl, wi -v is to extract the filter series weight from WeightMatl to perform the vertical scaling process, and the equations floor() and r〇Und() are respectively the flooring process and the rounding process. The decimal is over two and fetches the $=5 index and sets this index to the current vertical index y for the use of Xitian μ for selling vertical T#, rounding process 撷I247|2^~ Japanese repair (4) is replacing the page No.: 093111793 On July 12, 1994, the decimal result of this calculation was corrected and rounded to the nearest integer. In step 620, a security check is performed to confirm whether wi-v is Η. If YES, the index wi-v will exceed the reasonable range of 〇-1, and thus the weighted value cannot be generated. If wi-v = H, then go to step 630, reset the Wi-v value to 〇 and process the current vertical index y to perform the vertical placement process. If wi__v is not equal to H, then skip to step 64 and confirm the vertical index based on the input original image and the number of filter stages. In this κ % example, the number of filter stages is 4, so there are 4 vertical index indicators (匕vertical index p〇inter) iO, il, i2, and i3, and the four vertical index temples are respectively based on boundary conditions. The setting, the boundary condition is based on the following equation: "P + abs(pO)] = ((x + p)*e + he + (([υ - " ((-l)'a), where iU ] is corrected for the data of the level n, the range of n is between 〇 and 丨, the W is the length of the data, and the other parameters are listed below·· n= P + abs(p〇), P〇- - (L/2-1),

Pi = L/2, a two sign(x+p), b = sign((W-1)-(x+p)), c = WlX, d = x+p-(Wl), e = Not( b) & 1, 0 or 1 S The picture is the control flag. Figure 7 and Figure 8 points, when S箄; Ming different boundary conditions, Tian Temple in 1 series represents the choice of Figure 7, one, with the Tian s specializes in the representative of the Department of Figure 8

1247246 ... I July 12, 1994 Amendment — Case No.: 093111793 Option 2. The index p is from p〇 to the pi loop, the index X is self-twisted to the W-1 loop, and the equation sign(X) is as follows: sign{X) 1, X<0; 0, X >= ο. Equation abs(X) To calculate the absolute value of X, the equation Not(X) is a bit-wise operation that converts the bits of the binary, such as Not(O) = 1 and Not(l) = 〇. Note that although a and b here are the same as a and b in Fig. 3, the meanings are different. The vertical index index is generated according to the current vertical index of the boundary condition of selecting one. In this embodiment, if y=0, then i〇=y, il=y, i2=y+i, and i3=y+2 If y=H〇id - 2, then i〇=yi, n=:y, i2=:y+i and i3=y+i, if y=H〇id - 1, then i〇=y—i , ii=y, i2=y, and i3=y—i, otherwise 'iO=y_l, il=y, i2=y+l, and i3=y+2. After all the vertical index indicators have been determined, step 64 ends and jumps to step 65, which sets χ2 to zero. The flow chart of Figure 6 is connected to connection B of Figure 9 and is only used to continue the presentation of the entire process. The connection F of Figure 6 represents the flow back from Figure 9 to perform the flow chart of Figure 6, such as the operation of the partial loop. The process of vertical scanning of the original image to generate the vertical index indicator is over. Referring to Figure IX, Figure 9 is a flow chart of the horizontal scanning process. This flow chart is performed by connecting B to Figure 6. The horizontal scan is the same as the vertical scan, so the detailed description can refer to the vertical scan process.

Sh 氺χ2 x=floor(xl) wi-h=round[(xl-χ)χΗ] 12 Case pain. 093111793 July 12, 1994 Corrected where Sh is the horizontal scaling factor and x2 is the new image after scaling The horizontal index, χ1 is the old horizontal index of the input original image corresponding to x2, and the system is the integer part of χ1, and wi_h is the index of the filter series weighted from WeightMat2 to perform the horizontal scaling process. The horizontal index is generated according to the current horizontal index of the boundary condition. In this embodiment, if x=0, then i〇=x, il=x, i2=x+l, and i3=x+2, if x= W〇id - 2, then i〇=xl, il=x, i2=x+l and i3=x+l, if x=W〇ld - 1, then i〇=x-1, ii=x, i2 ==x, and i3=x-i, otherwise 'iO=xl, il=x, i2=x+l, and i3=x+2. After all of the horizontal index indicators are determined, step 73 ends and the horizontal scan ends. The flow chart of Figure 9 is connected to the connection C of Figure 10 and is only used to continue the presentation of the entire process. The process of horizontal scanning of the original image to generate the horizontal index indicator is over. Referring to FIG. 10, FIG. 10 is a flow chart of an engine for performing a scaling technique according to the present invention. Step 810: The segment data in the first block according to the vertical and horizontal index indicators is transmitted to the block buffer buffer for processing. In this embodiment, the number of vertical and horizontal filter stages is 4, therefore, the size of the buffer should be a minimum of 4χ4, which is equivalent to ^6 pixel units (pixebu un(1), if each pixel is a one-tuple, block, the size of the punch is 16-bit The block buffer β can be implemented in many different ways according to system requirements, but the general buffer memory includes DRM, sdram, flash memory, and the aforementioned memory inside the DSC, and includes a flip-flop. The register register (register-file, RF), and the aforementioned temporary storage in 1C. In addition, the 2D block buffer can borrow a 1D linear buffer and use the appropriate index side f to simulate 2D. The block buffer, based on the vertical index indicator and the horizontal index indicator, passes the block poor block to the block buffer B. The matrix data of the block buffer is as follows: B(〇,0)=SRC(j〇,i0) B (l,0)^SRC(jl,i〇) B(2, 0)=SRC( j2, i〇) 13 Case number: 09 3111793 Revised July 12, 1994 B(3,0)=SRC(j3, iO) B(0,1)=SRC(jO, il) B(l,1)-SRC(jl, il) B(2 , 1) = SRC (j2, i1) B (3, l) = SRC (j3, il) B (0, 2 > SRC (j0, i2) B (l, 2) = SRC (jl, i2) B ( 2,2)=SRC(j2,i2) B(3, 2)=SRC(j3, i2) B(0,3)=SRC(j0,i3)B(l,3)=SRC(jl,i3) B(2,3)=SRC(j2,i3) B(3,3)=SRC(j3, i3)

In step 830, the values of the fertilizer weights (f i Iter weighting) converted from WeightMatl and WeightMat2, respectively, and using wi_v and Wi_h to generate two vectors Wv and Wh °WV are generated according to the following equation:

Wv(0)=WeightMatl(wi_v, 0) Wv(l)=WeightMatl(wi_v, 1) Wv(2)=WeightMatl(wi_v, 2) Wv(3)=WeightMatl(wi_v, 3)

The value of Wh is generated according to the following formula:

Wh(0)=WeightMat2(wi_v, 〇) Wh(l)=WeightMat2(wi_v, 1) Wh(2)=WeightMat2(wi_v, 2) Wh(3)=WeightMat2(wi_v, 3)

After the corresponding data input block buffer β and vector l and the generation, step 850 performs matrix multiplication operation on the data in the block buffer B and the two vectors Wv and Wh, and Wv and the block buffer B. The data is used for inner product multiplication and also with Wh for inner product multiplication. The result of this matrix multiplication is a two-dimensional result, which is stored in the target buffer DST. The following mathematical expression shows the target buffer DST 14 1247246 (Case number: 093111793 The value in the July 12, 1994 correction is how DST(x2,y2) = Wv · B is obtained from the data of the block buffer. Wh _5(0,0)5(1,0)5(2,0)5(3,0)_ wh(°y (9) Wv (1) M;v(2) Wv(3)l· 5(0,1 5(1,1) 5(2,1) 5(3,1) 5(0,2) 5(1,2) 5(2,2) 5(3,2) 5(0,3) 5 (1,3) 5(2,3) 5(3,3) • W〆1) wh(2) where DST(x2,y2) is the output picture memory area, and DST(x2,y2) is also A pixel value is interpolated between the coordinates (x2, y2). A checking process is performed to confirm whether the scaling process has been completed. This checking process confirms whether the new horizontal index x2 is less than in step 870, and confirms the new in step 890. If the vertical index y2 is smaller than Hnew, if X2 is less than Wnew, jump to step 860 to increase X2 by 1 (χ2 = χ2 + 1) and return to step 710. Similarly, if y2 is less than Hnew, skip to step 880 and add y2 to I ( Y2=y2+1) and returning to step 610 of FIG. 6. Referring to FIG. 11, FIG. 11 is a schematic diagram of a zooming device 1 of the present invention, and the zooming device 1000 includes a raw buffer. 1010, for storing digital pictures, a processing unit 1015 for generating two weighting matrices, an image dividing unit (image divis〇r) 1020 for filtering digital images to generate a plurality of digital image blocks, one block The buffer 1030 is configured to continuously store the digital image block generated by the image dividing unit 1〇2〇, and a weighting matrix buffer for storing the weighting matrix, and a target buffer 1040 for storing the scaled Digital picture.

Please refer to FIG. 12, which is a schematic diagram of the processing unit 1〇15 in FIG. 11 for performing the matrix multiplication operation of the present invention. The processing unit 1〇15 includes a complex number: a multiplier and an adder electrically connected to the block buffer 1〇3〇, and the data unit of each block buffer 1030 is electrically connected to a multiplier, the multiplier The system is electrically connected to the receiving end Wh. The first row data unit (丨=〇) in block buffer B is multiplied by ^(〇), and the second row data unit (i = 1) in block buffer B is multiplied by Wh(1). Block Buffer Z 15 1247246 顺丨4)4 ΐ· 12... J ' Case No.: 093111793 The fourth row of data units (i = 2) multiplied by Wh(2) and blocks in July 12, 1994 The fourth row of data elements (i = 3) in buffer B is multiplied by Wh(3), and the result of these multiplication operations is electrically connected to an adder before being input to another multiplier and multiplied by Wv. The result of the multiplication operation is added by the adder in a column manner, such as the first column data unit in block buffer B (b 0) multiplied by Wv(0), and the second in block buffer B The column data unit (j = l) is multiplied by Wv(l), the third column data unit (j=2) in block buffer B is multiplied by Wv(2), and the fourth column in block buffer B is The data unit (j=3) is multiplied by Wv(3). Compared with the prior art, the present invention can greatly reduce the buffer memory requirement even when a digital image is enlarged, and further, the next block is continued after the vertical and horizontal scaling processes of each block are completed. Scaling, so the invention can be applied in real-time applications. The invention can also be applied to pictures of a single color as well as pictures of color. The above is only the preferred embodiment of the present invention, and the equivalent variations and modifications made by the present invention are intended to cover the scope of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS Fig. 1 is a conceptual diagram of a conventional enlargement process. Figure 2 is a conceptual diagram of the zooming technique of the present invention. Figure 3 is a flow chart for establishing a weighting matrix. Figure 4 is a flow chart of the zooming technique of the present invention. Figure 5 is a flow chart of initialization. Figure 6 is a flow chart of the vertical scanning process. Figure 7 and Figure 8 are schematic views of different boundary conditions of the present invention, respectively. Figure IX is a flow chart of the horizontal scanning process. 16 1247246 month day repair (test) is replacing page case number: 093111793 July 12, 1994 revision Figure 10 is a flow chart of the engine for performing the zooming technique of the present invention. Figure 11 is a schematic view of the zooming device of the present invention. Figure 12 is a schematic diagram of the processing unit of Figure 11. Symbol description of the schema 1000 Scaling device 1010 Raw buffer 1015 Processing unit 1020 Image splitting unit 1030 Block buffer 1035 Interpolated wave buffer 1040 Target buffer 17

Claims (1)

1247246 〇4 7. 12' Case No.: 093111793 July 12, 1994 曰Revised, patentable scope: 1. A method of scaling a digital image, the method comprising: (a) inputting the digital image; (b) according to The digital picture generates a block of the digital picture; (c) generating a first weighting matrix and a second weighting matrix; and (d) multiplying the block of the digital picture by the first The weighting matrix is multiplied by the second weighting matrix. 2. The method of claim 1, wherein the step (b) comprises interpolating the digital picture with an interpolation filter to generate a block of the digital picture. 3. The method of claim 2, wherein the block of the digital image generated in the step (b) is stored in a block buffer. 4. The method of claim 1, wherein the step (a) comprises inputting the digital picture into a source buffer. 5. The method of claim 1, wherein the step (c) comprises, according to the number of filter taps, sampling precision, and an adjusted scaling factor. The first weighting matrix and the second weighting matrix are generated. 6. The method of claim 5, wherein the adjustable scaling factor is a scaling factor, a preset value, a first weighting factor, a product of the scaling factor, or a second weighting factor. 18 J247246 rj Case No.: 093111793 Revised July 12, 1994. The method described in item 1 of the scope, further includes the “Pixel index” and the provision of boundary conditions before the execution of step (4). The method of item i, wherein the step (4) comprises: (e) multiplying the block of the digital picture by the first weighting matrix to generate an intermediate data; and (multiplying the intermediate tribute by the second weighting) a matrix to generate an output material., • The method of claim 8, wherein the method further comprises: storing the intermediate data in an intermediate buffer; and storing the output data in a target buffer. The method of item 9, wherein if it is still necessary to generate more output data from the block in step (4), repeat steps (4) and (1). L kinds of zoom-digital picture zooming device, the basin contains: - original, rushed a region for storing the digital image; a single unit for generating a first weighting matrix and a second weighting moment gate buffer for storing the first weighting matrix multiplied by the block of the digital image And the output data obtained by storing the data of the second weighted buffer. 12. The scaling device according to claim 11, wherein the processing unit comprises The first multiplier, using the number; to multiply each of the data in the block by the system 19 1247246, '一 / • a' Case number: 093111793 July 12, 94 revised a number of first adders, Each adder is coupled to a set of first multipliers for summing all results of the set of first multipliers; a plurality of second multipliers, each multiplier being coupled to a first adder, And multiplying the result of the first adder by a coefficient; and a second adder connected to the plurality of second multipliers for summing up the results of the plurality of second multipliers. The scaling device of claim 11, further comprising an interpolation filter for filtering the digital image to generate a block of the digital image. 14. The scaling as described in claim 13 Device, another area The buffer is connected to the interpolation filter for storing the block of the digital image. 20 -· ··~,•'.d;· :·*.Γ,•Name 12 even 6 匕! :警94.12 Case No.: 093111793 Revised on July 12, 1994, Fig. = 21 1247246 % : ...~ ' Case No.: 093111793 Revised 7, designated representative on July 12, 1994 Figure: (1) The representative representative figure of this case is: (11). (2) The symbol of the representative figure of the representative figure is simple: 1000 scaling device 1010 original buffer 1015 processing unit 1020 image segmentation unit 1030 block buffer 1035 Interpolation filter buffer 1040 Target buffer 捌, if there is a chemical formula in this case, please reveal the chemistry that best shows the characteristics of the invention.