CN100389591C - Method for carrying out zoom on two-dimensional digital signal - Google Patents
Method for carrying out zoom on two-dimensional digital signal Download PDFInfo
- Publication number
- CN100389591C CN100389591C CNB2006100116713A CN200610011671A CN100389591C CN 100389591 C CN100389591 C CN 100389591C CN B2006100116713 A CNB2006100116713 A CN B2006100116713A CN 200610011671 A CN200610011671 A CN 200610011671A CN 100389591 C CN100389591 C CN 100389591C
- Authority
- CN
- China
- Prior art keywords
- pixel
- digital signal
- dimensional digital
- vectors
- fifty
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Expired - Fee Related
Links
- 238000000034 method Methods 0.000 title claims abstract description 89
- 239000013598 vector Substances 0.000 claims abstract description 117
- 238000012952 Resampling Methods 0.000 claims description 37
- 230000008676 import Effects 0.000 claims description 12
- 230000000977 initiatory effect Effects 0.000 claims description 12
- 238000001914 filtration Methods 0.000 description 7
- 238000005070 sampling Methods 0.000 description 6
- 238000004364 calculation method Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 5
- 238000004422 calculation algorithm Methods 0.000 description 4
- 238000000205 computational method Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 230000000630 rising effect Effects 0.000 description 3
- 230000011218 segmentation Effects 0.000 description 2
- 230000003321 amplification Effects 0.000 description 1
- 125000004122 cyclic group Chemical group 0.000 description 1
- 238000009795 derivation Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 239000004744 fabric Substances 0.000 description 1
- 238000003780 insertion Methods 0.000 description 1
- 230000037431 insertion Effects 0.000 description 1
- 238000003199 nucleic acid amplification method Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Abstract
The present invention relates to a method for carrying out zoom on two-dimensional digital signals, which is used for reducing the two-dimensional digital signals of O[IO, JO], 0< =IO< W1 and 0< =JO< H1 of input W1*H1 on the longitudinal direction, and for inputting the two-dimensional digital signals of R[IR, JR], 0< =IR< W2, 0< =JR< H2 and H2< H1 of output W2*H2 line by line into row data in the O, namely, a pixel sequence of which the length is W1. The pixel sequence of which the length is W2 obtained from the resample is used as a pixel vector. The obtained S1 pixel vectors are uniformly divided by X1 times; the obtained S2 pixel vectors are uniformly divided by X2 times; two-dimensional digital signals R are obtained from one row of the data used as output two-dimensional digital signals R, wherein X1=H1%H2, X2=(H2-X1), S1=Ceil (H1/H2) and S2=floor (h1/h2). The amplifying method on the longitudinal direction is similar to the method. The method of the present invention can reduce the storage requirements for zooming the two-dimensional digital signals and the needed calculating amount of a zooming process.
Description
Technical field
The present invention relates to a kind of method of two-dimensional digital signal being carried out convergent-divergent.
Background technology
All be to realize the zoom technology of two-dimensional digital signal of the prior art (as two-dimensional digital signal) by filter.From digital signal processing theory, be length that the pixel sequences O resampling of W1 is that the pixel sequences R of length W2 requires to have following operation:
1: is length that the pixel of W1 is inserted 0 and handled and become the pixel S that length is W.W is the least common multiple of W1 and W2, W/W1=N1, W/W2=N2.This operation requires N1-1 null value of (or back) insertion before each original pixel;
2: S is carried out obtaining S ' after the Filtering Processing F disposal.The bandwidth of filter F is 2 π/N, and N=max (N1, N2);
3: S ' is carried out extraction operation obtain R, promptly extract a point every N2 point.
The filter that adopts when step 2 is generally the FIR filter, and iir filter can bring the advantage on internal memory and the stability relatively.Suppose that F is the FIR filter, its length is L, because F is a low pass filter, in order to satisfy its bandwidth condition, L is generally the integer more than or equal to 2N.
The image O of one width of cloth W1*H1 is being scaled in the image process of W2*H2 in real time, was using the method for low pass filter generally to need in the past by following steps (row-by-row system);
Initial I=J=0
1: the pixel O of delegation (I) among the input O, and I=I+1;
2: use low-pass filtering FW to be scaled the delegation pixel OR[I of width] as W2 to O (I);
3: OR[I] be stored among the capable BufferB and (delegation's pixel of sequence number minimum among the B override), judge whether the new line number of storing among the B enough carries out vertical filtering, if can not return step 1;
4: to the pixel rows OR[I that stores among the B], OR[I-1] ... vertical filtering of using filter FH to pursue row, the pixel of every row filtering output is formed the new pixel R[J of delegation], J=J+1;
This method faces following problem in practice:
1:FW all need adopt different filter coefficients could guarantee the quality of image R under different zoom ratio with FH.Just need design different FW and FH when needing different zoom ratio, and its coefficient is stored in the device, cause the demand of device storage capacity higher;
2: in theory, when FH be LH rank FIR filter the time, in Buffer B, need to preserve the intermediate object program OR of Ceil (LH/H2) row.The big more filter effect of LH is good more, if H=max (H1, H2), LH often H 2-4 doubly or more, cause the intermediate object program of the needs storage multirow of B;
3: this algorithm all uses filter on vertical and horizontal.Reaching good effect often needs the filter of high-order, and the exponent number of the amount of calculation of this method and filter is directly proportional, so amount of calculation is bigger.
Summary of the invention
The technical problem to be solved in the present invention provides a kind ofly carries out the method for convergent-divergent to two-dimensional digital signal, can reduce storage demand and the desired amount of calculation of convergent-divergent process.
In order to solve the problems of the technologies described above, the invention provides and a kind of two-dimensional digital signal is carried out the method for convergent-divergent, be used for the two-dimensional digital signal O[IO of W1*H1 that will input, JO], 0＜=IO＜W1,0＜=JO＜H1 dwindles in the vertical, the two-dimensional digital signal R[IR of output W2*H2, JR], 0＜=IR＜W2,0＜=JR＜H2, and H2＜H1 may further comprise the steps:
Import the line data among the two-dimensional digital signal O line by line, be that length is the pixel sequences of W1, the length that will obtain its resampling is that the pixel sequences of W2 is as a pixel vectors, have and X1 time S1 the pixel vectors that obtains be averaged, data line as output two-dimensional digital signal R, have and X2 time S2 the pixel vectors that obtains be averaged, data line as output two-dimensional digital signal R, obtain two-dimensional digital signal R, X1=H1%H2 wherein, X2=(H2-X1), S1=Ceil (H1/H2), S2=Floor (H1/H2), Ceil () represents to round up computing, Floor () represents downward rounding operation, " % " expression complementation computing, H1=S1*X1+S2*X2.
Further, said method also can have following characteristics: also operation that is averaged by S1 pixel vectors for X1 time and the operation that is averaged by S2 pixel vectors for X2 time will be scatter fifty-fifty.
Further, said method also can have following characteristics: be to realize in the two-dimensional digital signal convergent-divergent process that by following steps the operation that will be averaged by S1 pixel vectors for X1 time and the operation that is averaged by S2 pixel vectors for X2 time are scattered fifty-fifty:
(a) initiation parameter: D=is less than any natural number or 0 of H2; C=0; JO=JR=0; A[n]=B[n]=0, n=0 ... W2-1;
(b) judge whether JO＞H1 sets up, if finish, otherwise carry out next step;
(c) the data line O[m of input two-dimensional digital signal O, JO], m=0 ... W1-1; Resampling is the pixel vectors B[n that comprises W2 pixel], n=0 ... W2-1, and make JO=JO+1;
(d) make A[n]=A[n]+B[n], B[n]=0, n=0 ... W2-1; D=D+H2; C=C+1;
(e) if D＜H1 turns back to step (b); Otherwise carry out next step;
(f) obtain the data line R[n of the two-dimensional digital signal R that will export, JR]=A[n]/C, A[n]=0, n=0 ... W2-1; C=0; D=D-H1; JR=JR+1 returns step (b).
Further, said method also can have following characteristics: be to realize in the two-dimensional digital signal convergent-divergent process that by following steps the operation that will be averaged by S1 pixel vectors for X1 time and the operation that is averaged by S2 pixel vectors for X2 time are scattered fifty-fifty:
(a) initiation parameter, S1=Ceil (H1/H2), S2=Floor (H1/H2); Deta=H1-H2*S2; D=is arbitrarily less than the natural number or 0 of H2; LT=S1 or S2; T=0; A[n]=B[n]=0, n=0 ... W2-1; JO=JR=0;
(b) judge whether JO＞H1 sets up, if finish, otherwise carry out next step;
(c) length is the line data O[m of W1 among the input two-dimensional digital signal O, JO], m=0 ... W1-1; Resampling is the pixel vectors B[n that comprises W2 pixel], n=0 ... W2-1, and make JO=JO+1;
(d) make A[n]=A[n]+B[n], B[n]=0, n=0 ... W2-1; T=T+1;
(e) if T＜LT returns step (b); Otherwise, carry out next step;
(f) obtain the data line of the two-dimensional digital signal R that will export: R[n, JR]=A[n]/LT, n=0 ... W2-1; T=0; A[0]=A[1]=... A[W2-1]=0; JR=JR+1;
(g) make D=D+Deta,, return step (b) if D＜H2 makes LT=S2; Otherwise make D=D-H2, LT=S1 returns step (b).
Further, said method also can have following characteristics: the line data in the two-dimensional digital signal O that will import line by line, and promptly length is the pixel sequences of W1, resample and be pixel vectors B, be length when being the pixel sequences of W2, if W1＞W2 then carries out by the following method:
Progressively importing length is W1 pixel of the pixel sequences of W1, there is the inferior individual pixel of S1 ' of X1 ' to be averaged a pixel as pixel vectors B to input, there is the inferior individual pixel of S2 ' of X2 ' to be averaged a pixel as pixel vectors B to input, thereby obtain pixel vectors B, and above-mentioned operation that is averaged by the individual pixel of S1 ' and the operation that is averaged by the individual pixel of S2 ' are scatter fifty-fifty, X1 '=W1%W2 wherein, X2 '=(W2-X1 '), S1 '=Ceil (W1/W2), S2 '=Floor (W1/W2), Ceil () represents to round up computing, Floor () represents downward rounding operation, " % " expression complementation computing, W1=S1 ' * X1 '+S2 ' * X2 ';
If W2＞W1 then carries out by the following method:
When progressively importing length and be W1 pixel of pixel sequences of W1, the individual pixel of S1 ' that has the inferior pixel will importing of X1 ' to copy as pixel vectors B, the individual pixel of S2 ' that has the inferior pixel will importing of X2 ' to copy as pixel vectors B, thereby obtain pixel vectors B, and inferior copy operation of above-mentioned X1 ' and the inferior copy operation of X2 ' are scatter fifty-fifty, X1 '=W2%W1 wherein, X2 '=(W1-X1 '), S1 '=Ceil (W2/W1), S2 '=Floor (W2/W1), Ceil () represents to round up computing, Floor () represents downward rounding operation, " % " expression complementation computing, W2=S1 ' * X1 '+S2 ' * X2 '.
Further, said method also can have following characteristics: described two-dimensional digital signal is a digital picture.
In order to solve the problems of the technologies described above, the present invention provides a kind of again two-dimensional digital signal has been carried out the method for convergent-divergent, is used for the two-dimensional digital signal O[IO of W1*H1 that will input, JO], 0＜=IO＜W1,0＜=JO＜H1 amplifies in the vertical, the two-dimensional digital signal R[IR of output W2*H2, JR], 0＜=IR＜W2,0＜=JR＜H2, and H1＜H2 may further comprise the steps:
Import the line data among the two-dimensional digital signal O line by line, be that length is the pixel sequences of W1, the length that will obtain its resampling is that the pixel sequences of W2 is as a pixel vectors, there is a pixel vectors that will obtain for X1 time to copy as S1 the pixel vectors of output two-dimensional digital signal R, there is a pixel vectors that will obtain for X2 time to copy as S2 the pixel vectors of output two-dimensional digital signal R, obtain two-dimensional digital signal R, X1=H2%H1 wherein, X2=(H1-X1), S1=Ceil (H2/H1), S2=Floor (H2/H1), Ceil () represents to round up computing, Floor () represents downward rounding operation, " % " expression complementation computing, H2=S1*X1+S2*X2.
Further, said method also can have following characteristics: also a pixel vectors that X1 time resampling is obtained will be copied as the operation of S1 pixel vectors of output two-dimensional digital signal R and the operation that a pixel vectors that X2 time obtains resampling copies as S2 pixel vectors exporting two-dimensional digital signal R and scatter fifty-fifty.
Further, said method also can have following characteristics: be to realize in the two-dimensional digital signal convergent-divergent process above-mentioned X1 pixel vectors copy operation and X2 pixel vectors copy operation is scattered fifty-fifty by following steps:
(a) initiation parameter: D=is less than any natural number or 0 of H1; A[0]=A[1] ...=A[W2-1]=0; JO=JR=0;
(b), finish, otherwise carry out next step if JO＞H1 has obtained exporting two-dimensional digital signal;
(c) length is the line data O[m of W1 among the input two-dimensional digital signal O, JO], m=0 ... W1-1; Resampling is the pixel vectors A[n that comprises W2 pixel], n=0 ... W2-1; Make JO=JO+1;
(d) make R[n, JR]=A[n], n=0 ... W2-1; D=D+H1; JR=JR+1;
(e) if D＜H2 returns step (d), otherwise, make D=D-H2, return step (b).
Further, said method also can have following characteristics: be to realize in the two-dimensional digital signal convergent-divergent process above-mentioned X1 pixel vectors copy operation and X2 pixel vectors copy operation is scattered fifty-fifty by following steps:
(a) initiation parameter: S1=Ceil (H2/H1), S2=Floor (H2/H1); Deta=H2%H1; D=is arbitrarily less than the natural number or 0 of H1; LT=S1 or S2; T=0; A[0]=A[1] ...=A[W2-1]=0; JO=JR=0;
(b) if JO＞H1 finishes, otherwise carries out next step;
(c) length is the line data O[m of W1 among the input two-dimensional digital signal O, JO], m=0 ... W1-1; Resampling is the pixel vectors A[n that comprises W2 pixel], n=0 ... W2-1; Make JO=JO+1;
(d) make R[n, JR]=A[n], n=0 ... W2-1; T=T+1; JR=JR+1
(e) if T＜LT returns step (d), otherwise, making D=D+Deta, T=0 carries out next step;
(f) if D＞=H2 makes D=D-H2, LT=S1 returns step (b), otherwise makes LT=S2, returns step (b).
Further, said method also can have following characteristics: the line data in the two-dimensional digital signal O that will import line by line, and promptly length is the pixel sequences of W1, resample and be pixel vectors A, be length when being the pixel sequences of W2, if W1＞W2 then carries out by the following method:
Progressively importing length is W1 pixel of the pixel sequences of W1, there is the inferior individual pixel of S1 ' of X1 ' to be averaged a pixel as pixel vectors A to input, there is the inferior individual pixel of S2 ' of X2 ' to be averaged a pixel as pixel vectors A to input, thereby obtain pixel vectors A, and above-mentioned operation that is averaged by the individual pixel of S1 ' and the operation that is averaged by the individual pixel of S2 ' are scatter fifty-fifty, X1 '=W1%W2 wherein, X2 '=(W2-X1 '), S1 '=Ceil (W1/W2), S2 '=Floor (W1/W2), Ceil () represents to round up computing, Floor () represents downward rounding operation, " % " expression complementation computing, W1=S1 ' * X1 '+S2 ' * X2 ';
If W2＞W1 then carries out by the following method:
When progressively importing length and be W1 pixel of pixel sequences of W1, the individual pixel of S1 ' that has the inferior pixel will importing of X1 ' to copy as pixel vectors A, the individual pixel of S2 ' that has the inferior pixel will importing of X2 ' to copy as pixel vectors A, thereby obtain pixel vectors A, and inferior copy operation of above-mentioned X1 ' and the inferior copy operation of X2 ' are scatter fifty-fifty, X1 '=W2%W1 wherein, X2 '=(W1-X1 '), S1 '=Ceil (W2/W1), S2 '=Floor (W2/W1), Ceil () represents to round up computing, Floor () represents downward rounding operation, " % " expression complementation computing, W2=S1 ' * X1 '+S2 ' * X2 '.
Further, said method also can have following characteristics: described two-dimensional digital signal is a digital picture.
In sum, adopt the inventive method, avoided in the prior art, the problem that different zoom ratio need adopt different filters to carry out filtering has greatly been saved memory space ground demand, has also saved the desired amount of calculation of convergent-divergent process simultaneously.
Description of drawings
Fig. 1 is the embodiment of the invention to the flow chart of the two-dimensional digital signal convergent-divergent and the method for dwindling in the vertical.
Fig. 2 is the embodiment of the invention to the flow chart of the two-dimensional digital signal convergent-divergent and the method for amplifying in the vertical.
Embodiment
In the present invention, be in real time be length the image R that the image O of W1*H1 is scaled W2*H2 in real time.W1, H1, W2, H2 is any natural number, and laterally zoom ratio is W2/W1, and vertically zoom ratio is H2/H1.Two-dimensional digital signal O can carry out convergent-divergent line by line or by row ground.
The present invention, resamples to be divided into down-sampled again and to rise and samples based on the method that pixel sequences is resampled the convergent-divergent of image.
In the present embodiment, be length that the pixel sequences O resampling of W1 is the pixel sequences R of length W2, establish this resampling for down-sampled, i.e. W2＜W1.Its method is as follows:
If S1 '=Ceil (W1/W2), S2 '=Floor (W1/W2), wherein Ceil () represents to round up computing, Floor () represents downward rounding operation.
Definition again: X1 '=W1%W2, X2 '=(W2-X1 '), i.e. X1 '+X2 ' W2." % " expression complementation computing.Significantly, have:
W1＝S1’*X1’+S2’*X2’，
The implication of above-mentioned derivation is the pixel for the W1 among the O, is divided into the W2 section, wherein has X1 ' section to contain the individual pixel of S1 ', has X2 ' section to contain the individual pixel of S2 '.If we ask the average pixel of this section of average back output just to obtain W2 pixel just at last to the picture element of each section wherein, realized the process that from O to R, must resample.
But, if have only segmentation in this process and ask average operation, the partial distortion that can cause pixel sequences O, because certain part among the O is by 1 picture element of the individual picture element average out to of S1 ', in addition-a little partly by 1 picture element of the individual picture element average out to of S2 ', though S1 ' and S2 ' are relatively more approaching but also unequal, cause this two-part resolution inconsistent.Therefore, must find a kind of method that these two kinds of different parts of resolution are scatter fifty-fifty, see on the whole that promptly the part of two kinds of different resolutions is scattered among the sequence R very fifty-fifty, so just can avoid R distortion visually.
Suppose W1 '=27, W2 '=8, S1 '=Ceil (W1/W2)=4 then, S2 '=Floor (W1/W2)=3, X1 '=W1%W2=3, X2 '=(W2-X1)=5.Have:
W1=S1 ' * X1 '+S2 ' * X2 '=4*3+3*5=27, it is on average to be obtained by the picture element among continuous 4 O that 3 points are promptly arranged in the R sequence, it is on average to be obtained by the picture element among continuous 3 O that 5 points are arranged.For fear of visual distortion, 34 picture elements that become 1 and 53 change 1 should on average spread out on the whole.Otherwise may allow the observer aware image fault.
When below introduce a kind of down-sampled concrete grammar, be used for control and when the individual point of S1 ' asked on average, be that the individual point of S2 ' is asked on average, and two kinds of different resolutions are scatter.
Import the pixel sequences O that length is W1 one by one, each picture element is expressed as O[I], 0＜=I＜W1 exports the pixel sequences R that length is W2 one by one, each picture element is expressed as R[J], 0＜=J＜W2.Carry out following computing in the resampling process:
Step 110, initializing variable D '=arbitrarily is less than W2 and more than or equal to 0 integer, A '=0; C '=0; I=J=0;
Step 120 if I, has obtained the pixel sequences R that will export more than or equal to W1, finishes; Otherwise carry out next step;
Step 130 is from the new pixel O[I of sequence O input], make I progressively increase 1, A '=A '+O[I], D '=D '+W2; C '=C '+1;
Step 140, if D '＜W1 returns step 120, otherwise, carry out next step;
Step 150 makes R[J]=A '/C ', C '=0, J=J+1, D '=D '-W1, step 120 is returned in A '=0.
Though this flow process does not directly calculate S1 ' and S2 ', the iteration control by to variables D has produced and will ask average operation and the individual input picture element of S2 ' is asked the average average scattered effect of operation the individual input picture element of S1 '.For W1=27, the example of W2=8, D ' is initialized as 2, and then operation result is successively 4,3,3,4,3,3,4,3 picture elements to be asked all, as seen evenly scatter.
For the scattered fifty-fifty down-sampled method of part that resolution is different with two kinds, top method is not unique, for example, can also adopt following method:
Steps A A calculates S1 '=Ceil (W1/W2), S2 '=Floor (W1/W2), and increment deta '=W1%W2, initialization LT '=S2 ' (or S1 '), cyclic variable J=0 is arbitrarily less than W2 and more than or equal to an integer D ' of 0;
Step BB takes out the individual pixel of LT ' from sequence O, this LT ' pixel is got average, as the picture element R[J of the pixel sequences R that will export], make J=J+1 then;
Step CC judges whether J equals W2, if, obtained the pixel sequences R that will export, finish, otherwise, carry out next step;
Step DD makes D '=D '+deta ', whether judges D ' more than or equal to W2, if, execution in step EE, otherwise, execution in step FF;
Step e E makes D '=D '-W2, and LT '=S1 ' returns step BB;
Step F F makes LT '=S2 ', returns step BB.
For W1=27, the example of W2=8, S1 '=4, S2 '=3, X1 '=3, X2 '=5 make deta '=3, initialization D '=4, LT '=3, then operation result is successively 3,3,4,3,4,3,3,4 picture elements to be asked all, also evenly scatter.
Will be length that the pixel sequences O resampling of W1 is the pixel sequences R of length W2, and this resample to rising sampling, i.e. W2＞W1, its method is as follows:
With down-sampled similar, make S1 '=Ceil (W2/W1), S2 '=Floor (W2/W1).X1 '=W2%W1, X2 '=(W1-X1 '), then have: W2=S1 ' * X1 '+S2 ' * X2 '.
During resampling, for the picture element among the O, wherein the individual pixel of X1 ' duplicates the individual pixel of the inferior S1 of becoming ' of S1 ' by a pixel, and the individual pixel of X2 ' duplicates the individual pixel of the inferior S2 of becoming ' of S2 ' by a pixel in addition, obtains the pixel sequences R that length is W2 ' at last.Similarly, must guarantee two kinds differently resolution can on average spread out on the whole, could guarantee that the pixel sequences R that obtains is visually undistorted.
Suppose W1=8, W2=27, S1 '=Ceil (W2/W2)=4 then, S2 '=Floor (W2/W2)=3, X1 '=W2%W1=3, X2 '=(W1-X1)=5.Have
W2=S1 ' * X1 '+S2 ' * X2 '=4*3+3*5=27, promptly the picture element among the O has 5 to duplicate 3 times, has 3 to duplicate 4 times, just can obtain the sequence R of 27 picture elements.The pixel that duplicates 3 times must scatter with the pixel that duplicates 4 times, otherwise will cause distortion.
When introduce a kind of concrete grammar that rises sampling below, be used for control and when the individual point of S1 ' duplicated, be that the individual point of S2 ' duplicates, and two kinds of different resolutions are scatter.
Import the pixel sequences O that length is W1 one by one, each picture element is expressed as O[I], 0＜=I＜W1 exports the pixel sequences R that length is W2 one by one, each picture element is expressed as R[J], 0＜=J＜W2.This resamples to rising sampling, i.e. W2＞W1.In the resampling process, carry out following steps:
Step 210, initializing variable D '=arbitrarily is less than W1 and more than or equal to 0 integer, I=J=0;
Step 220 if I, has obtained the pixel sequences R that will export more than or equal to W1, finishes; Otherwise, carry out next step;
Step 230 is from the new pixel O[I of sequence O input], make I=I+1;
Step 240 makes R[J]=O[I], J=J+1; D '=D '+W1;
Step 250, if D '＜W2 returns step 240, otherwise, make D '=D '-W2, return step 220.
Though this flow process does not directly calculate S1 ' and S2 ', but by to variables D ' iteration control, produced inferior operation and the inferior scattered fifty-fifty effect of operation that pixel of sequence O is copied as the individual output sequence R of S2 ' pixel of X2 ' that pixel of sequence O is copied as the individual output sequence R of S1 ' pixel of X1 '.For W1=8, the example of W2=27, D ' is initialized as 5, and then operation result is successively each input pixel to be duplicated 3,4,3,3,4,3,3,4 times, as seen evenly scatter.
For the scattered fifty-fifty method of sampling that rises of part that resolution is different with two kinds, top method neither be unique, for example, can also adopt following method:
Step HH calculates S1 '=Ceil (W2/W1) earlier, and S2 '=Floor (W2/W1), initializing variable deta '=W2%W1, D ' are arbitrarily less than W1 and more than or equal to 0 integer, LT '=S2 ' (or S1 '), I=J=0;
Step II if I＞=W1 has obtained the pixel sequences R that will export, finishes, otherwise carries out next step;
Step JJ, the new pixel O[I of input from sequence W1], make I=I+1;
Step K K, O[J]=R[I], J=J+1; And it is inferior to repeat this operation LT ';
Step LL makes D '=D '+deta ', if D '＞=W1 then makes D '=D '-W1, LT '=S1 ' returns Step II; Otherwise make LT '=S2 ', return Step II.
Present embodiment is in that (wherein each pixel is expressed as O[IO the two-dimensional digital signal O of a width of cloth W1*H1, JO], 0＜=IO＜W1,0＜=JO＜H1) be scaled W2*H2 two-dimensional digital signal R (wherein each pixel is expressed as R[IR, JR], 0＜=IR＜W2, during 0＜=JR＜H2), need be to each line data among the two-dimensional digital signal O, promptly length is that the pixel sequences of W1 resamples (rise or fall), the pixel sequences that to obtain H1 length be W2.At this moment, if is the length that obtains that the pixel sequences of W2 is regarded a pixel vectors as, then can be that the pixel sequences of W2 regards that vertically to go up length be the pixel vectors sequence O ' of H1 as with this H1 length, from here as can be seen, as long as the pixel value computing in the above method for resampling is replaced with the pixel vectors computing, just can finish pixel vectors sequence O ' with same method for resampling is the pixel vectors sequence R of H2, i.e. the convergent-divergent computing of two-dimensional digital signal R to length.Certainly, in the actual operation process,, carry out to the resampling process of the line data of two-dimensional digital signal O with to the resampling process that vertically goes up pixel vectors sequence O ' is staggered for the save memory space.
Below still according to H1＞H2 or two kinds of situations of H2＞H1, promptly vertically dwindle or the situation of amplifying describes, but it should be noted that, the horizontal and vertical some dimensions that are not fixed as two-dimensional digital signal herein, but can exchange.
To be length that the pixel vectors sequence O ' of H1 resamples to length is the pixel vectors sequence R of H2, resample to down-sampled as this, i.e. H2＜H1.Its method is as follows:
Definition: S1=Ceil (H1/H2), S2=Floor (H1/H2), X1=H1%H2, X2=(H2-X1), i.e. X1+X2=H2.Wherein Ceil () represents to round up computing, and Floor () represents downward rounding operation, " % " expression complementation computing.Have:
H1＝S1*X1+S2*X2，
When resampling, with the pixel vectors of the H1 among the O ', be divided into the H2 section, wherein there is the X1 section to contain S1 pixel vectors, there is the X2 section to contain S2 pixel vectors.If we ask the average pixel vector of this section of average back output to the pixel vectors of each section wherein, just obtain H2 pixel vectors just at last, realized resampling process from O ' to R.To asking of a plurality of pixel vectors all is exactly that the pixel on the same position in a plurality of pixel vectors is asked all pixels of this position of back conduct output pixel vectors.
Similarly, if have only segmentation in this process and ask average operation, the partial distortion that can cause pixel vectors sequence O ', because certain part among the O ' is by 1 pixel vectors of S1 pixel vectors average out to, the other part causes this two-part resolution inconsistent by 1 pixel vectors of S2 pixel vectors average out to.Therefore, also to scatter by the part average mark that these two kinds of resolution are different, avoid distortion visually.Its method also is similar to pixel value.
Will be length that the pixel vectors sequence O ' resampling of H1 is the pixel vectors sequence R of length H2, and this resample to rising sampling, i.e. H1＜H2, its method is as follows:
Definition: S1=Ceil (H2/H1), S2=Floor (H2/H1), X1=H2%H1, X2=(H1-X1) then has: H2=S1*X1+S2*X2.
Promptly when resampling, for the picture element among the O ', wherein X1 pixel vectors duplicated by a pixel vectors and becomes S1 pixel vectors for S1 time, and X2 pixel vectors duplicated by a pixel vectors and become S2 pixel vectors for S2 time in addition, and obtaining length at last is the pixel vectors sequence R of H2.Similarly, must guarantee two kinds differently resolution can on average spread out on the whole, could guarantee that the pixel vectors sequence R that obtains is visually undistorted.
A kind of to two-dimensional digital signal convergent-divergent and the concrete computational methods of dwindling in the vertical as shown in Figure 1, may further comprise the steps:
Import the two-dimensional digital signal O[IO of W1*H1 line by line, JO], 0＜=IO＜W1,0＜=JO＜H1, export the two-dimensional digital signal R[IR of W2*H2 line by line, JR], 0＜=IR＜W2,0＜=JR＜H2, H2＜H1 vertically dwindles the following computing of execution in the process at image, as shown in Figure 1:
Step 320 judges whether JO＞=H1 sets up, if, obtained exporting two-dimensional digital signal, finish, otherwise carry out next step;
Step 340 makes A[n]=A[n]+B[n], B[n]=0, n=0 ... W2-1; D=D+H2; C=C+1;
Step 350 is if D＜H1 turns back to step 320; Otherwise carry out next step;
May find out from top flow process, if delegation's dateout that resampling is obtained as a pixel vectors, then the vertically process of dwindling of this digital image is exactly the down-sampled process to this pixel vectors sequence, its flow process and similar.
Another kind of concrete computational methods are as follows:
Step 2 judges whether JO＞=H1 sets up, if, obtained exporting two-dimensional digital signal, finish, otherwise carry out next step;
Step 3, length is the line data O[m of W1 among the input two-dimensional digital signal O, JO], m=0 ... W1-1; Resampling is the pixel vectors B[n that comprises W2 pixel], n=0 ... W2-1, and make JO=JO+1;
Step 4 makes A[n]=A[n]+B[n], B[n]=0, n=0 ... W2-1; T=T+1;
Step 5 judges whether T＜LT sets up, if return step 2; Otherwise, carry out next step;
Step 6 obtains the data line among the two-dimensional digital signal R: R[n, JR]=A[n]/LT, n=0 ... W2-1; T=0; A[0]=A[1]=... A[W2-1]=0; JR=JR+1;
Step 7 makes D=D+Deta, judges whether D＜H2 sets up, if, make LT=S2, return step 2, otherwise make D=D-H2; LT=S1 returns step 2.
A kind of to the two-dimensional digital signal convergent-divergent and the concrete computational methods of amplifying in the vertical as shown in Figure 2, may further comprise the steps:
Import the two-dimensional digital signal O[IO of W1*H1 line by line, JO], 0＜=IO＜W1,0＜=JO＜H1, export the two-dimensional digital signal R[IR of W2*H2 line by line, JR], 0＜=IR＜W2,0＜=JR＜H2, H2＞H1 carries out following computing in the vertical amplification process of image, as shown in Figure 2:
Step 420 judges whether JO＞=H1 sets up, if, obtained exporting two-dimensional digital signal, finish, otherwise carry out next step;
Step 440 makes R[n, JR]=A[n], n=0 ... W2-1; D=D+H1; JR=JR+1;
Step 450 judges whether D＜H2 sets up, if set up, returns step 440, otherwise, make D=D-H2, return step 420.
Similarly, another kind of concrete computational methods are as follows:
Step 1 ', initiation parameter: S1=Ceil (H2/H1), S2=Floor (H2/H1); Deta=H2%H1; D=is arbitrarily less than the natural number or 0 of H1; LT=S1 or S2; T=0; A[0]=A[1] ...=A[W2-1]=0; JO=JR=0;
Step 2 ', judge whether JO＞=H1 sets up, if, obtained exporting two-dimensional digital signal, finish, otherwise carry out next step;
Step 3 ', length is the line data O[m of W1 among the input picture O, JO], m=0 ... W1-1; Resampling is the pixel vectors A[n that comprises W2 pixel], n=0 ... W2-1; Make JO=JO+1;
Step 4 ', make R[n, JR]=A[n], n=0 ... W2-1; T=T+1; JR=JR+1
Step 5 ', if T＜LT returns step 4 ', otherwise, making D=D+Deta, T=0 carries out next step;
Step 6 ', if D＞=H1 makes D=D-H1, LT=S1 returns step 2 ', otherwise make LT=S2, return step 2 '.
From top several algorithms as can be seen, the actual buffer memory that uses only requires that length is the buffer memory A of W2 except control variables, simultaneously, compare with the method for using low-pass filtering in the past, avoided using a large amount of add operations of taking advantage of, only need a spot of addition and divide operations.
More than provide some specific algorithms, but the present invention is not limited to this,, average scattered other algorithm of different resolution part can be suitable for as long as in the resampling process of pixel sequences or pixel vectors sequence.
Especially, in another embodiment, also can adopt other method for resampling to finish the liter sampling of pixel sequences or down-sampled, also only adopt the method for the foregoing description to finish resampling the pixel sequence vector.Has the effect that reduces amount of calculation and save memory equally.
Apparent you easily see, though embodiment is example with the digital picture, the present invention handles applicable to the convergent-divergent of described two-dimensional digital signal.
Claims (12)
1. method of two-dimensional digital signal being carried out convergent-divergent, be used for two-dimensional digital signal O[IO with the W1*H1 of input, JO], 0＜=IO＜W1,0＜=JO＜H1 dwindles in the vertical, the two-dimensional digital signal R[IR of output W2*H2, JR], 0＜=IR＜W2,0＜=JR＜H2, and H2＜H1 may further comprise the steps:
Import the line data among the two-dimensional digital signal O line by line, be that length is the pixel sequences of W1, the length that will obtain its resampling is that the pixel sequences of W2 is as a pixel vectors, have and X1 time S1 the pixel vectors that obtains be averaged, data line as output two-dimensional digital signal R, have and X2 time S2 the pixel vectors that obtains be averaged, data line as output two-dimensional digital signal R, obtain two-dimensional digital signal R, X1=H1%H2 wherein, X2=(H2-X1), S1=Ceil (H1/H2), S2=Floor (H1/H2), Ceil () represents to round up computing, Floor () represents downward rounding operation, " % " expression complementation computing, H1=S1*X1+S2*X2.
2. the method for claim 1 is characterized in that, also operation that is averaged by S1 pixel vectors for X1 time and the operation that is averaged by S2 pixel vectors for X2 time will be scatter fifty-fifty.
3. method as claimed in claim 2 is characterized in that, is to realize in the two-dimensional digital signal convergent-divergent process that by following steps the operation that will be averaged by S1 pixel vectors for X1 time and the operation that is averaged by S2 pixel vectors for X2 time are scattered fifty-fifty:
(a) initiation parameter: D=is less than any natural number or 0 of H2; C=0; JO=JR=0; A[n]=B[n]=0, n=0 ... W2-1;
(b) judge whether JO＞H1 sets up, if finish, otherwise carry out next step;
(c) the data line O[m of input two-dimensional digital signal O, JO], m=0 ... W1-1; Resampling is the pixel vectors B[n that comprises W2 pixel], n=0 ... W2-1, and make JO=JO+1;
(d) make A[n]=A[n]+B[n], B[n]=0, n=0 ... W2-1; D=D+H2; C=C+1;
(e) if D＜H1 turns back to step (b); Otherwise carry out next step;
(f) obtain the data line R[n of the two-dimensional digital signal R that will export, JR]=A[n]/C, A[n]=0, n=0 ... W2-1; C=0; D=D-H1; JR=JR+1 returns step (b).
4. method as claimed in claim 2 is characterized in that, is to realize in the two-dimensional digital signal convergent-divergent process that by following steps the operation that will be averaged by S1 pixel vectors for X1 time and the operation that is averaged by S2 pixel vectors for X2 time are scattered fifty-fifty:
(a) initiation parameter, S1=Ceil (H1/H2), S2=Floor (H1/H2); Deta=H1-H2*S2; D=is arbitrarily less than the natural number or 0 of H2; LT=S1 or S2; T=0; A[n]=B[n]=0, n=0 ... W2-1; JO=JR=0;
(b) judge whether JO＞H1 sets up, if finish, otherwise carry out next step;
(c) length is the line data O[m of W1 among the input two-dimensional digital signal O, JO], m=0 ... W1-1; Resampling is the pixel vectors B[n that comprises W2 pixel], n=0 ... W2-1, and make JO=JO+1;
(d) make A[n]=A[n]+B[n], B[n]=0, n=0 ... W2-1; T=T+1;
(e) if T＜LT returns step (b); Otherwise, carry out next step;
(f) obtain the data line of the two-dimensional digital signal R that will export: R[n, JR]=A[n]/LT, n=0 ... W2-1; T=0; A[0]=A[1]=... A[W2-1]=0; JR=JR+1;
(g) make D=D+Deta,, return step (b) if D＜H2 makes LT=S2; Otherwise make D=D-H2, LT=S1 returns step (b).
5. the method for claim 1 is characterized in that, the line data in the two-dimensional digital signal O that will import line by line, be that length is the pixel sequences of W1, resampling is pixel vectors B, when promptly length is the pixel sequences of W2, if W1＞W2 then carries out by the following method:
Progressively importing length is W1 pixel of the pixel sequences of W1, there is the inferior individual pixel of S1 ' of X1 ' to be averaged a pixel as pixel vectors B to input, there is the inferior individual pixel of S2 ' of X2 ' to be averaged a pixel as pixel vectors B to input, thereby obtain pixel vectors B, and above-mentioned operation that is averaged by the individual pixel of S1 ' and the operation that is averaged by the individual pixel of S2 ' are scatter fifty-fifty, X1 '=W1%W2 wherein, X2 '=(W2-X1 '), S1 '=Ceil (W1/W2), S2 '=Floor (W1/W2), Ceil () represents to round up computing, Floor () represents downward rounding operation, " % " expression complementation computing, W1=S1 ' * X1 '+S2 ' * X2 ';
If W2＞W1 then carries out by the following method:
When progressively importing length and be W1 pixel of pixel sequences of W1, the individual pixel of S1 ' that has the inferior pixel will importing of X1 ' to copy as pixel vectors B, the individual pixel of S2 ' that has the inferior pixel will importing of X2 ' to copy as pixel vectors B, thereby obtain pixel vectors B, and inferior copy operation of above-mentioned X1 ' and the inferior copy operation of X2 ' are scatter fifty-fifty, X1 '=W2%W1 wherein, X2 '=(W1-X1 '), S1 '=Ceil (W2/W1), S2 '=Floor (W2/W1), Ceil () represents to round up computing, Floor () represents downward rounding operation, " % " expression complementation computing, W2=S1 ' * X1 '+S2 ' * X2 '.
6. the method for claim 1 is characterized in that, described two-dimensional digital signal is a digital picture.
7. method of two-dimensional digital signal being carried out convergent-divergent, be used for two-dimensional digital signal O[IO with the W1*H1 of input, JO], 0＜=IO＜W1,0＜=JO＜H1 amplifies in the vertical, the two-dimensional digital signal R[IR of output W2*H2, JR], 0＜=IR＜W2,0＜=JR＜H2, and H1＜H2 may further comprise the steps:
Import the line data among the two-dimensional digital signal O line by line, be that length is the pixel sequences of W1, the length that will obtain its resampling is that the pixel sequences of W2 is as a pixel vectors, there is a pixel vectors that will obtain for X1 time to copy as S1 the pixel vectors of output two-dimensional digital signal R, there is a pixel vectors that will obtain for X2 time to copy as S2 the pixel vectors of output two-dimensional digital signal R, obtain two-dimensional digital signal R, X1=H2%H1 wherein, X2=(H1-X1), S1=Ceil (H2/H1), S2=Floor (H2/H1), Ceil () represents to round up computing, Floor () represents downward rounding operation, " % " expression complementation computing, H2=S1*X1+S2*X2.
8. method as claimed in claim 7, it is characterized in that, also a pixel vectors that X1 time resampling is obtained will be copied as the operation of S1 pixel vectors of output two-dimensional digital signal R and the operation that a pixel vectors that X2 time obtains resampling copies as S2 pixel vectors exporting two-dimensional digital signal R and scatter fifty-fifty.
9. method as claimed in claim 8 is characterized in that, is to realize in the two-dimensional digital signal convergent-divergent process above-mentioned X1 pixel vectors copy operation and X2 pixel vectors copy operation is scattered fifty-fifty by following steps:
(a) initiation parameter: D=is less than any natural number or 0 of H1; A[0]=A[1] ...=A[W2-1]=0; JO=JR=0;
(b), finish, otherwise carry out next step if JO＞H1 has obtained exporting two-dimensional digital signal;
(c) length is the line data O[m of W1 among the input two-dimensional digital signal O, JO], m=0 ... W1-1; Resampling is the pixel vectors A[n that comprises W2 pixel], n=0 ... W2-1; Make JO=JO+1;
(d) make R[n, JR]=A[n], n=0 ... W2-1; D=D+H1; JP=JR+1;
(e) if D＜H2 returns step (d), otherwise, make D=D-H2, return step (b).
10. method as claimed in claim 8 is characterized in that, is to realize in the two-dimensional digital signal convergent-divergent process above-mentioned X1 pixel vectors copy operation and X2 pixel vectors copy operation is scattered fifty-fifty by following steps:
(a) initiation parameter: S1=Ceil (H2/H1), S2=Floor (H2/H1); Deta=H2%H1; D=is arbitrarily less than the natural number or 0 of H1; LT=S1 or S2; T=0; A[0]=A[1] ...=A[W2-1]=0; JO=JR=0;
(b) if JO＞H1 finishes, otherwise carries out next step;
(c) length is the line data O[m of W1 among the input two-dimensional digital signal O, JO], m=0 ... W1-1; Resampling is the pixel vectors A[n that comprises W2 pixel], n=0 ... W2-1; Make JO=JO+1;
(d) make R[n, JR]=A[n], n=0 ... W2-1; T=T+1; JR=JR+1;
(e) if T＜LT returns step (d), otherwise, making D=D+Deta, T=0 carries out next step;
(f) if D＞=H1 makes D=D-H1, LT=S1 returns step (b), otherwise makes LT=S2, returns step (b).
11. method as claimed in claim 7 is characterized in that, the line data in the two-dimensional digital signal O that will import line by line, be that length is the pixel sequences of W1, resampling is pixel vectors A, when promptly length is the pixel sequences of W2, if W1＞W2 then carries out by the following method:
Progressively importing length is W1 pixel of the pixel sequences of W1, there is the inferior individual pixel of S1 ' of X1 ' to be averaged a pixel as pixel vectors A to input, there is the inferior individual pixel of S2 ' of X2 ' to be averaged a pixel as pixel vectors A to input, thereby obtain pixel vectors A, and above-mentioned operation that is averaged by the individual pixel of S1 ' and the operation that is averaged by the individual pixel of S2 ' are scatter fifty-fifty, X1 '=W1%W2 wherein, X2 '=(W2-X1 '), S1 '=Ceil (W1/W2), S2 '=Floor (W1/W2), Ceil () represents to round up computing, Floor () represents downward rounding operation, " % " expression complementation computing, W1=S1 ' * X1 '+S2 ' * X2 ';
If W2＞W1 then carries out by the following method:
When progressively importing length and be W1 pixel of pixel sequences of W1, the individual pixel of S1 ' that has the inferior pixel will importing of X1 ' to copy as pixel vectors A, the individual pixel of S2 ' that has the inferior pixel will importing of X2 ' to copy as pixel vectors A, thereby obtain pixel vectors A, and inferior copy operation of above-mentioned X1 ' and the inferior copy operation of X2 ' are scatter fifty-fifty, X1 '=W2%W1 wherein, X2 '=(W1-X1 '), S1 '=Ceil (W2/W1), S2 '=Floor (W2/W1), Ceil () represents to round up computing, Floor () represents downward rounding operation, " % " expression complementation computing, W2=S1 ' * X1 '+S2 ' * X2 '.
12. method as claimed in claim 7 is characterized in that, described two-dimensional digital signal is a digital picture.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2006100116713A CN100389591C (en) | 2006-04-13 | 2006-04-13 | Method for carrying out zoom on two-dimensional digital signal |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CNB2006100116713A CN100389591C (en) | 2006-04-13 | 2006-04-13 | Method for carrying out zoom on two-dimensional digital signal |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN1829277A CN1829277A (en) | 2006-09-06 |
| CN100389591C true CN100389591C (en) | 2008-05-21 |
Family
ID=36947385
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CNB2006100116713A Expired - Fee Related CN100389591C (en) | 2006-04-13 | 2006-04-13 | Method for carrying out zoom on two-dimensional digital signal |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN100389591C (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO1991004636A2 (en) * | 1989-09-11 | 1991-04-04 | Independent Broadcasting Authority | Improvements in or relating to motion adaptive video signal converters |
| JP2003242497A (en) * | 2002-01-10 | 2003-08-29 | Ricoh Co Ltd | Low-resolution image creating method and device |
| JP2004229040A (en) * | 2003-01-24 | 2004-08-12 | Murata Mach Ltd | Image processing device |
| CN1607826A (en) * | 2003-10-16 | 2005-04-20 | 凌阳科技股份有限公司 | Directionality interpolative method and device for increasing image resolution |
-
2006
- 2006-04-13 CN CNB2006100116713A patent/CN100389591C/en not_active Expired - Fee Related
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO1991004636A2 (en) * | 1989-09-11 | 1991-04-04 | Independent Broadcasting Authority | Improvements in or relating to motion adaptive video signal converters |
| JP2003242497A (en) * | 2002-01-10 | 2003-08-29 | Ricoh Co Ltd | Low-resolution image creating method and device |
| JP2004229040A (en) * | 2003-01-24 | 2004-08-12 | Murata Mach Ltd | Image processing device |
| CN1607826A (en) * | 2003-10-16 | 2005-04-20 | 凌阳科技股份有限公司 | Directionality interpolative method and device for increasing image resolution |
Also Published As
| Publication number | Publication date |
|---|---|
| CN1829277A (en) | 2006-09-06 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| EP3120322B1 (en) | Method for processing input low-resolution (lr) image to output high-resolution (hr) image | |
| CN111429347A (en) | Image super-resolution reconstruction method and device and computer-readable storage medium | |
| CN112330542A (en) | Image reconstruction system and method based on CRCSAN network | |
| CN106709875A (en) | Compressed low-resolution image restoration method based on combined deep network | |
| CN107123094B (en) | Video denoising method mixing Poisson, Gaussian and impulse noise | |
| CN102881000A (en) | Super-resolution method, device and equipment for video image | |
| CN112801877A (en) | Super-resolution reconstruction method of video frame | |
| JP2009524861A (en) | Method and apparatus for improving resolution of digital image | |
| CN111652804B (en) | Super-resolution reconstruction method based on expansion convolution pyramid and bottleneck network | |
| CN111932461A (en) | Convolutional neural network-based self-learning image super-resolution reconstruction method and system | |
| CN108830792B (en) | Image super-resolution method using multi-class dictionary | |
| CN109785279B (en) | Image fusion reconstruction method based on deep learning | |
| CN108492252A (en) | Face image super-resolution reconstruction method based on secondary reconstruction | |
| CN108122262B (en) | Sparse representation single-frame image super-resolution reconstruction algorithm based on main structure separation | |
| CN114926336A (en) | Video super-resolution reconstruction method and device, computer equipment and storage medium | |
| CN112150356A (en) | Single compressed image super-resolution reconstruction method based on cascade framework | |
| CN100389591C (en) | Method for carrying out zoom on two-dimensional digital signal | |
| WO2008148207A1 (en) | Method and device for selecting optimal transform matrices for down-sampling dct image | |
| CN116188272B (en) | Two-stage depth network image super-resolution reconstruction method suitable for multiple fuzzy cores | |
| Lu et al. | Utilizing homotopy for single image superresolution | |
| US20120050820A1 (en) | Image processing apparatus, control method of the same, and program | |
| CN111476743B (en) | Digital signal filtering and image processing method based on fractional differential | |
| CA2687489A1 (en) | Method and device for down-sampling a dct image in the dct domain | |
| CN111179169B (en) | Parameter optimization method of bicubic image interpolation model | |
| CN1333372C (en) | Resampling method for digital signal |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20080521 Termination date: 20200413 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |