CN101511011A - Display method and device for image drop sampling quick decode - Google Patents
Display method and device for image drop sampling quick decode Download PDFInfo
- Publication number
- CN101511011A CN101511011A CNA2009100961991A CN200910096199A CN101511011A CN 101511011 A CN101511011 A CN 101511011A CN A2009100961991 A CNA2009100961991 A CN A2009100961991A CN 200910096199 A CN200910096199 A CN 200910096199A CN 101511011 A CN101511011 A CN 101511011A
- Authority
- CN
- China
- Prior art keywords
- dct
- sampled
- image
- piece
- unit
- 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.)
- Pending
Links
Images
Abstract
The invention discloses a fast image-down-sampling decoding and displaying method and includes the steps that firstly down-sampling proportions in the horizontal and vertical directions are confirmed according to resolutions of original images and images output set by users; secondly, bit streams of coding images are input and variable length decoding and inverse quantization operation of bit streams are carried out; thirdly, down-sampling operation of DCT fields of DCT blocks obtained after the inverse quantization is implemented according to down-sampling proportions in the horizontal and vertical directions in the first step; fourthly, the operation of inverse DCT transformation of the DCT blocks down-reduced is carried out and low-resolution images to be output are reconstructed; fifthly, reconstructed images decoded are output. The method can fast decode high-resolution images into low-resolution images with any down-sampling rates with lower calculation complexity and can better preserve the quality of images.
Description
Technical field
The present invention relates to image code domain, more particularly, the present invention relates to the method and apparatus that a kind of image drop sampling quick decode shows.
Background technology
At present, the demand of digital consumer electronics products increases surprising, and wherein imaging and video class product proportion are very big.High-resolution digital camera, digital photo frame with 4,000,000 to 1,000 ten thousand pixels, and the camera mobile phone of built in digital camera become very universal, and puts terminal equipment such as (as mp4) such as the personal video record with big capacity storage application program (PVR), family audio frequency and video center and portable multimedia broadcasting and can both carry out the thin function of taking a picture.
But lower, the limited feature of information fan-out capability (screen size etc.) of power that these equipment have large storage capacity processor (CPU), carry out the full detail that image can not output code flow when playing comprises, therefore must reduce the resolution of image according to the screen display size.For example, the high definition original image that 8,000,000 pixel digital cameras generate has 3400 * 2300 pixels, and computer monitor has the resolution of 1024 * 768 pixels usually, washes generally that 5 one-by-one inch photograph sheets are set 1200 * 840 pixels, flushing is set 2400 * 1920 pixels for 10 cun and just can have been satisfied fully.And picture material is developing to high-resolution direction now, and this causes the required amount of calculation of decoded picture code stream and memory space very big, and therefore, image drop sampling is more and more necessary.
At present, traditional image drop sampling method is to carry out down-sampled operation in spatial domain, at first the coded image code stream is all decoded, and utilizes the low pass filter technology that decoded code stream is carried out down-sampled operation then, code stream is encoded again at last again.This is a kind of very directly method, because it has increased the design of low pass filter, will certainly increase whole operand, and then increases the use of processor resource.Discrete cosine transform (Discrete Cosine Transform, i.e. dct transform) is a kind of mathematical operation that is closely related with Fourier transform, and IDCT is an inverse discrete cosine transform.The down-sampled technology in DCT territory is by realizing the down-sampled of image in the DCT territory, and do not need as the down-sampled method of traditional images, earlier image all to be decoded, carry out down-sampled again, encode again at last, greatly reduce computational complexity, improved the speed of image drop sampling decoding, and can keep picture quality preferably, thereby come into one's own.
As mentioned above, in traditional image drop sampling method, the image that utilizes low cpu power to generate, show and browse big capacity high definition in the short time be difficult, speed is slow excessively.Therefore, the fast decoding display packing of image drop sampling is provided, to reach the resolution of rapid adjustment image, one of focus that thereby generation of raising picture decoding and surfing are the insiders to be paid close attention to, and realize that by the down-sampled technology in DCT territory the fast decoding of image drop sampling also will be one of important directions of future development.
Summary of the invention
Defective or deficiency at above-mentioned prior art existence, the present invention proposes the method that a kind of image drop sampling quick decode shows, this method can generate the low-resolution image of big or small down-sampled rate arbitrarily according to customer requirements fast by the fast decoding of the down-sampled technology realization in DCT territory image.
To achieve these goals, the present invention adopts following technical solution:
A kind of image drop sampling quick decode display packing may further comprise the steps:
(1) input coding original image, the user sets the resolution of output image;
(2) resolution of the output image of setting according to the resolution of original image and user is determined the down-sampled ratio of horizontal direction and the down-sampled ratio of vertical direction;
If the resolution of original image is M * N, the resolution of the output image that the user sets is m * n, and the down-sampled rate of so described level is M/m, and vertical down-sampled rate is N/n, M wherein, and N, m, n is positive integer.
(3) the coded image code stream of input is carried out decoding length changeable code and inverse quantization operation;
If the DCT block size is M * N, the down-sampled ratio of horizontal direction is X
c: Y
c, the down-sampled ratio of vertical direction is X
r: Y
r, wherein M, N, X
c, Y
c, X
rAnd Y
rBe positive integer;
If satisfy M/ (X
c: Y
c), N/ (X
r: Y
r) be integer, be initial point then in starting point, x axle positive direction to the right, in the downward rectangular coordinate system of y axle positive direction, upper left corner decoding line promptly was the point (M/X-1 in the rectangular coordinate system, N/Y-1) and be parallel to the straight line of DCT piece counter-diagonal, decoding length changeable code is decoded to the zone that the upper left corner decoding line of DCT piece and x axle and y axle surround at least;
If do not satisfy M/ (X
c: Y
c), N/ (X
r: Y
r) be integer, decoding length changeable code then carries out the full decoder operation.
(4) according to the down-sampled ratio of level that obtains in the step (1) and vertical direction, the DCT piece that obtains behind the inverse quantization is carried out the down-sampled operation in DCT territory.
For width of cloth size is the original image of m * n, and establishing the DCT block size is M * N, and the down-sampled ratio of horizontal direction is X
c: Y
c, the down-sampled ratio of vertical direction is X
r: Y
r, wherein M, N, X
c, Y
c, X
rAnd Y
rBe positive integer.
In order to improve the decoding speed of image drop sampling as far as possible, in the down-sampled operation in DCT territory, judge, can adopt the different down-sampled methods in DCT territory for different down-sampled ratios.
If satisfy M/ (X
c: Y
c), N/ (X
r: Y
r) be integer, then adopt direct intercepting method that the DCT piece is carried out the down-sampled operation in DCT territory.
Described direct intercepting method is upper left corner M/ (Xc:Yc) * N/ (Xr:Yr) zone of each the DCT coefficient block behind the direct intercepting inverse quantization, then the DCT inverse transformation is carried out in this zone.
If do not satisfy M/ (X
c: Y
c), N/ (X
r: Y
r) be integer, then adopt reflection method or piece method for splitting that the DCT piece is carried out the down-sampled operation in DCT territory.
Described employing reflection method carries out the down-sampled operation in DCT territory to the DCT piece, promptly is to extract size successively by matrixing from original image to be (M * (X
c: Y
c)) * (N * (X
r: Y
r)) the zone, then dct transform is carried out in this zone, the DCT inverse transformation is carried out to this zone then in the intercepting DCT piece zone of size down-sampled output image afterwards for the low frequency part of M * N is wherein again.
Described employing piece method for splitting carries out the down-sampled operation in DCT territory to the DCT piece, at first to X in the original image
c* X
rEach DCT piece of individual DCT piece carries out the DCT inverse transformation of M * N, obtains the big or small (M * X of being in the pixel domain
c) * (N * X
r) block of pixels, this block of pixels is carried out (M * X
c) * (N * X
r) dct transform, then the intercepting wherein size be (M * Y
c) * (N * Y
r) low frequency DCT coefficient region, again the DCT inverse transformation is carried out in this low frequency coefficient zone.
When the down-sampled ratio of level greater than M and be that integral multiple and the vertical down-sampled ratio of M is greater than N and when being the integral multiple of N, can at first directly intercept the DC component (DC) in each DCT piece upper left corner behind the inverse quantization, then the DC after the intercepting being reconfigured to new size is the DCT piece of M * N, adopts direct intercepting method to carry out the down-sampled operation in DCT territory to this piece again; In order to prevent frequency alias, when adopting said method, before image output, can add a filter.
When the down-sampled ratio of level greater than M but satisfies is that integral multiple and the vertical down-sampled ratio of M is greater than N but do not satisfy when being the integral multiple of N, can at first directly intercept the DC component (DC) in each DCT piece upper left corner behind the inverse quantization, then the DC after the intercepting being reconfigured to new size is the DCT piece of M * N, adopts reflection method or piece Split Method to carry out the down-sampled operation in DCT territory to this piece again.
In addition, the down-sampled operation in DCT territory also can be adopted other down-sampled methods in DCT territory, the DCT territory convolution method of mentioning in for example many documents, when level is m/n with vertical down-sampled rate, this method is at first carried out the liter sampling of n:1 to the DCT piece that obtains from code stream, and then carry out the down-sampled technology of m:1, thereby finish the image drop sampling of n/m arbitrary proportion.But this method operand is very big, requires also high to memory space.
(5) the DCT piece after down-sampled is carried out the operation of DCT inverse transformation, rebuild low-resolution image to be exported, obtain reconstructed image;
(6) reconstructed image behind the output decoder.
A kind of image drop sampling quick decode display unit, it comprises with the lower part:
The coded image code stream input unit that is used for the image input;
The image drop sampling quick decode device that is used for the down-sampled decoding of image space;
Be used for the display unit that decoded image shows;
The image drop sampling quick decode device is connected between coded image code stream input unit and the display unit;
Described image drop sampling quick decode device comprises:
Be used for decoding the variable length decoding unit of the DCT piece of process quantification from image;
The DCT piece that is used for quantizing carries out the inverse quantization unit of inverse quantization;
Be used to need to judge whether the determining device 1 of part decoding length changeable code;
Be used to control the controller of decoding length changeable code;
Be used to select the determining device 2 of the down-sampled unit of DCT;
Be used for directly intercepting the down-sampled unit of integer DCT of method;
Be used to realize the down-sampled unit of DCT of the down-sampled method of other DCT;
The DCT inverse transformation unit that is used for inverse transformation.
The input of described determining device 1 links to each other with the output of described coded image code stream input unit; The output of described determining device 1 links to each other with the input of described variable length decoding unit; The output of described variable length decoding unit links to each other with the input of described inverse quantization unit; The output of described inverse quantization unit links to each other with the input of described determining device 2; The output of described inverse transformation unit links to each other with the input of display unit; Down-sampled unit of described integer DCT and the down-sampled unit of described DCT are connected in parallel between determining device 2 and the inverse transformation unit; Described controller is connected between determining device 1 and the variable length decoding unit.
The present invention directly handles the DCT piece that the I two field picture of jpeg image or MPEG obtains in decode procedure, the down-sampled technology in DCT territory is by realizing the down-sampled of image in the DCT territory, avoided earlier image all being decoded in the traditional image drop sampling method, carry out down-sampled again, last process of encoding again, greatly reduce computational complexity, improved the speed of image drop sampling decoding, and can keep picture quality preferably, thereby can reach the treatment effect that adopts the low pass filter technology, decoding speed that again can the accelerogram picture.
Description of drawings
Fig. 1 is the flow chart of image drop sampling quick decode display packing of the present invention;
Fig. 2 is the flow chart of JPEG (joint photographic experts group) (JPEG) picture decoding process;
Fig. 3 is the concrete flow chart of implementing of the inventive method;
Seven kinds of template schematic diagrames of direct intercepting method during Fig. 4 implements for the present invention is concrete;
DC component during Fig. 5 implements for the present invention is concrete is reassembled as the schematic diagram of new DCT piece;
The schematic diagram of reflection method during Fig. 6 implements for the present invention is concrete;
The schematic diagram of piece Split Method during Fig. 7 implements for the present invention is concrete;
Fig. 8 is the functional block diagram of image drop sampling quick decode display unit of the present invention.
Embodiment
Below in conjunction with accompanying drawing a kind of image drop sampling quick decode display packing of the present invention is elaborated by embodiment.
As shown in Figure 1, a kind of image drop sampling quick decode display packing may further comprise the steps:
A kind of image drop sampling quick decode display packing may further comprise the steps:
(1) input coding original image, the user sets the resolution of output image;
(2) resolution of the output image of setting according to the resolution of original image and user is determined the down-sampled ratio of horizontal direction and the down-sampled ratio of vertical direction;
If the resolution of original image is M * N, the publish picture resolution of picture of the speed that the user sets is m * n, and the down-sampled rate of so described level is M/m, and vertical down-sampled rate is N/n, M wherein, and N, m, n are positive integer.
(3) the coded image code stream of input is carried out decoding length changeable code and inverse quantization operation;
If the DCT block size is M * N, the down-sampled ratio of horizontal direction is X
c: Y
c, the down-sampled ratio of vertical direction is X
r: Y
r, wherein M, N, X
c, Y
c, X
rAnd Y
rBe positive integer;
If satisfy M/ (X
c: Y
c), N/ (X
r: Y
r) be integer, be initial point then in starting point, x axle positive direction to the right, in the downward rectangular coordinate system of y axle positive direction, upper left corner decoding line promptly was the point (M/X-1 in the rectangular coordinate system, N/Y-1) and be parallel to the straight line of DCT piece counter-diagonal, decoding length changeable code is decoded to the zone that the upper left corner decoding line of DCT piece and x axle and y axle surround at least;
If do not satisfy M/ (X
c: Y
c), N/ (X
r: Y
r) be integer, decoding length changeable code then carries out the full decoder operation.
(4) according to the down-sampled ratio of level that obtains in the step (1) and vertical direction, the DCT piece that obtains behind the inverse quantization is carried out the down-sampled operation in DCT territory.
For width of cloth size is the original image of m * n, and establishing the DCT block size is M * N, and the down-sampled ratio of horizontal direction is X
c: Y
c, the down-sampled ratio of vertical direction is X
r: Y
r, wherein M, N, X
c, Y
c, X
rAnd Y
rBe positive integer.
In order to improve the decoding speed of image drop sampling as far as possible, in the down-sampled operation in DCT territory, judge, can adopt the different down-sampled methods in DCT territory for different down-sampled ratios.
If satisfy M/ (X
c: Y
c), N/ (X
r: Y
r) be integer, then adopt direct intercepting method that the DCT piece is carried out the down-sampled operation in DCT territory.
Described direct intercepting method is upper left corner M/ (Xc:Yc) * N/ (Xr:Yr) zone of each the DCT coefficient block behind the direct intercepting inverse quantization, then the DCT inverse transformation is carried out in this zone.For example, for piece image, wherein the size of DCT piece is 8 * 8, if the down-sampled rate of horizontal direction and vertical direction all is 2:1,4 * 4 scopes in the upper left corner are exactly desired low frequency DCT coefficient region so, if the down-sampled rate of horizontal direction and vertical direction all is 4:1,2 * 2 scopes in the upper left corner are exactly desired DCT coefficient region so.
If do not satisfy M/ (X
c: Y
c), N/ (X
r: Y
r) be integer, then adopt reflection method or piece method for splitting that the DCT piece is carried out the down-sampled operation in DCT territory.
Described employing reflection method carries out the down-sampled operation in DCT territory to the DCT piece, promptly is to extract size by matrixing from original image to be (M * (X
c: Y
c)) * (N * (X
r: Y
r)) the zone, then dct transform is carried out in this zone, intercepting is the DCT piece zone of size down-sampled output image afterwards for the low frequency part of M * N is wherein.For example, suppose output image behind the down-sampling by the macro block of M ' * N ' individual 16 * 16 (macro block comprises 48 * 8 piece, and dct transform is to be the unit with 8 * 8 piece) composition, wherein M ' represents vertical respectively and macro block width horizontal direction with N '; Pairing original picture is made up of M * N macro block, and wherein M and N are respectively the macro block width on vertical and the direction.So, for the down-sampled ratio R of level
xWith vertical down-sampled ratio R
y, R is arranged
x=M '/M, R
y=N '/N, R
xWith R
yMight be unequal.If choose minimum integer m and n, so that h=4mR
yAnd w=4nR
xBe integer, then the frame behind down-sampling just be divided into P * Q=(4M '/m) * (4N '/n) individual size is the piece of 4m * 4n, each piece corresponding again in the former image size be the mapping block of h * w.Therefore, original picture also can be split into P * Q size and is the mapping block of h * w.In addition, mapping block in the original image and DCT block boundary may not line up.
Described employing piece method for splitting carries out the down-sampled operation in DCT territory to the DCT piece, at first to X in the original image
c* X
rEach DCT piece of individual DCT piece carries out the DCT inverse transformation of M * N, obtains the big or small (M * X of being in the pixel domain
c) * (N * X
r) block of pixels, this block of pixels is carried out (M * X
c) * (N * X
r) dct transform, then the intercepting wherein size be (M * Y
c) * (N * Y
r) low frequency DCT coefficient region, again the DCT inverse transformation is carried out in this low frequency coefficient zone.For example, DCT piece for the m * m that obtains behind the inverse quantization individual 8 * 8, at first each DCT piece is carried out the DCT inverse transformation, obtain size and be the block of pixels of 8m * 8m, again this block of pixels is carried out the dct transform of 8m * 8m, the low frequency DCT coefficient part of intercepting 8n * 8n is wherein carried out the DCT inverse transformation to this low frequency part at last and is got final product then.
When the down-sampled ratio of level greater than M and be that integral multiple and the vertical down-sampled ratio of M is greater than N and when being the integral multiple of N, can at first directly intercept the DC component (DC) in each DCT piece upper left corner behind the inverse quantization, then the DC after the intercepting being reconfigured to new size is the DCT piece of M * N, adopts direct intercepting method to carry out the down-sampled operation in DCT territory to this piece again; In order to prevent frequency alias, when adopting said method, before image output, can add a filter.For example, for width of cloth size is the image of m * n, wherein the size of DCT piece is 8 * 8, if the down-sampled rate of horizontal direction and vertical direction all is 16:1, can at first directly intercept the DC component in the upper left corner in each DCT piece so, then these DC being reassembled into new size is 8 * 8 piece, again this piece is used 2 * 2 the zone in the direct intercepting method intercepting upper left corner, thereby realize that horizontal direction and the down-sampled rate of vertical direction all are 16:1, so just reduce operand, greatly improved the speed of picture decoding.
When the down-sampled ratio of level greater than M but satisfies is that integral multiple and the vertical down-sampled ratio of M is greater than N but do not satisfy when being the integral multiple of N, can at first directly intercept the DC component (DC) in each DCT piece upper left corner behind the inverse quantization, then the DC after the intercepting being reconfigured to new size is the DCT piece of M * N, adopts reflection method or piece Split Method to carry out the down-sampled operation in DCT territory to this piece again.For example, for width of cloth size is the image of m * n, wherein the size of DCT piece is 8 * 8, if the down-sampled rate of horizontal direction and vertical direction all is 15:1, can at first directly intercept the DC component in the upper left corner in each DCT piece so, then these DC being reassembled into new size is 8 * 8 piece, adopt reflection method to realize level and vertical down-sampled rate 15: 8 to this piece again with the piece Split Method, thereby realize that horizontal direction and the down-sampled rate of vertical direction all are 15:1, can improve the speed of picture decoding.
In addition, the down-sampled operation in DCT territory also can be adopted other down-sampled methods in DCT territory, the DCT territory convolution method of mentioning in for example many documents, when level is m/n with vertical down-sampled rate, this method is at first carried out the liter sampling of n:1 to the DCT piece that obtains from code stream, and then carry out the down-sampled technology of m:1, thereby finish the image drop sampling of n/m arbitrary proportion.But this method operand is very big, requires also high to memory space.
(4) the DCT piece after down-sampled is carried out the operation of DCT inverse transformation, rebuild low-resolution image to be exported, obtain reconstructed image;
(5) low-resolution image behind the output decoder.
As shown in Figure 8, a kind of image drop sampling quick decode display unit comprises with the lower part:
The coded image code stream input unit that is used for the image input;
The image drop sampling quick decode device that is used for the down-sampled decoding of image space;
Be used for the display unit that decoded image shows;
The image drop sampling quick decode device is connected between coded image code stream input unit and the display unit;
Described image drop sampling quick decode device comprises:
Be used for decoding the variable length decoding unit of the DCT piece of process quantification from image;
The DCT piece that is used for quantizing carries out the inverse quantization unit of inverse quantization;
Be used to need to judge whether the determining device 1 of part decoding length changeable code;
Be used to control the controller of decoding length changeable code;
Be used to select the determining device 2 of the down-sampled unit of DCT;
Be used for directly intercepting the down-sampled unit of integer DCT of method;
Be used to realize the down-sampled unit of DCT of the down-sampled method of other DCT;
The DCT inverse transformation unit that is used for inverse transformation.
The input of described determining device 1 links to each other with the output of described coded image code stream input unit; The output of described determining device 1 links to each other with the input of described variable length decoding unit; The output of described variable length decoding unit links to each other with the input of described inverse quantization unit; The output of described inverse quantization unit links to each other with the input of described determining device 2; The output of described inverse transformation unit links to each other with the input of display unit; Down-sampled unit of described integer DCT and the down-sampled unit of described DCT are connected in parallel between determining device 2 and the inverse transformation unit; Described controller is connected between determining device 1 and the variable length decoding unit.
The present invention can implement different images in the concrete process of using, effect for this method of simple declaration, be jpeg compressed image or MPEG I two field picture with coded image below, pending DCT piece is that 8 * 8 DCT piece is as exemplary enforcement.As shown in Figure 3, at first the resolution information of the output image of setting according to the resolution of original image and user is determined down-sampled rate of horizontal direction and the down-sampled rate of vertical direction (201); (202) are according to the original image of input and the information of output image then, judge whether to carry out the variable length code full decoder, establish N=(output image resolution sizes/original image resolution size) * 8 herein, (2031) if 1≤N≤7, N is an integer, then carry out local decoding length changeable code (2032), otherwise, variable length code full decoder (2033) carried out; Then be that (204) carry out inverse quantization operation to the code stream behind the decoding length changeable code, obtain the DCT piece; And then do a judgement, (2051) if 1≤N≤7, N is an integer, the DCT piece after then (2053) adopt direct intercepting method to inverse quantization is carried out the down-sampled operation in DCT territory, otherwise the DCT piece after then (2052) adopt reflection method or piece Split Method to inverse quantization carries out the down-sampled operation in DCT territory; (206) carry out the operation of DCT inverse transformation then, rebuild needed low-resolution image; Be the image of rebuilding behind (207) output decoder at last.
Following specific implementation process is elaborated to direct intercepting method, and the down-sampled rate of image level and vertical direction of establishing is 2:1.After DCT piece behind the inverse quantization used direct intercepting method, 4 * 4 zones, the upper left corner of DCT piece were the low frequency information of DCT piece.Fig. 4 is the down-sampled rate of level and vertical direction for less than 8 positive integer the time, seven kinds of template schematic diagrames, when output image be original image 1/8 the time, then intercept the DC component shown in the template A; When output image be original image 2/8 the time, then intercept the zone of 2 * 2 among the template B; When output image be original image 3/8 the time, then intercept the zone of 3 * 3 among the template C; When output image be original image 1/2 the time, then intercept the zone of 4 * 4 among the template D; When output image be original image 5/8 the time, then intercept the zone of 5 * 5 among the template E; When output image be original image 3/4 the time, then intercept the zone of 6 * 6 among the template F; When output image be original image 7/8 the time, then intercept the zone of 7 * 7 among the template G.Thereby, can generate the needed low-resolution image of quick generation, can keep the quality of image again preferably.
Wherein the template A among Fig. 4 is a kind of extremely special situation, promptly only need the DC component (DC) in the decoding DCT piece, to greatly reduce the operand in the picture decoding process like this, this special case can have important use in practice along with the improving constantly of digital product resolution.For example, the original image of one 8,000,000 pixel digital camera generation has 3400 * 2300 pixels, if the resolution of screen display has only CIF (352*288), so only needs the decoding DC coefficient just enough.If during the DC component of only decoding, the resolution of the image that generates still is higher than the resolution of screen display, at this time, just the DC component of 8 * 88 * 8 adjacent DCT piece (promptly be 64 * 64 DCT piece) can be formed new 8 * 8 DCT piece, as shown in Figure 5, then this DCT piece is carried out down-sampledly again, meet the low-resolution image of screen resolution with final generation.In addition, when the original image of ultrahigh resolution is only decoded the DC component (DC) of DCT piece,, therefore can before image output, carry out filtering operation if direct output image may cause spectral aliasing.
In order to realize that will have high-resolution original image fast decoding is the low-resolution image that meets the arbitrary proportion of user's request, the down-sampled operation in DCT territory can be taked the reflection method among the present invention, as shown in Figure 6.Below reflection method is elaborated, establishing coded image is jpeg compressed image or MPEG I two field picture, and pending DCT piece is 8 * 8 DCT piece.Suppose output image behind the down-sampling by the macro block of M ' * N ' individual 16 * 16 (macro block comprises 48 * 8 piece, and dct transform is that 8 * 8 piece is the unit) composition, wherein M ' represents vertical respectively and macro block width horizontal direction with N '; Pairing original picture is made up of M * N macro block, and wherein M and N are respectively the macro block width on vertical and the direction.So, for the down-sampled ratio R of level
xWith vertical down-sampled ratio R
y, R is arranged
x=M '/M, R
y=N '/N, R
xWith R
yMight be unequal.If choose minimum integer m and n, so that h=4mR
yAnd w=4nR
xBe integer, then the frame behind down-sampling just be divided into P * Q=(4M '/m) * (4N '/n) individual size is the piece of 4m * 4n, each piece corresponding again in the former image size be the mapping block of h * w.Therefore, original picture also can be split into P * Q size and is the mapping block of h * w.As can be seen, mapping block in the original image and DCT block boundary may not line up yet from Fig. 6.
Below by a concrete example reflection method is further described, described coded image is jpeg compressed image or MPEG I two field picture, the size of macro block is 16 * 16, and pending DCT piece is 8 * 8 DCT piece, and level is 3:2 with vertical down-sampled ratio.For a width of cloth size is 384 * 288 original image, and the width of level and vertical macroblocks is respectively 24 and 18; The level of output image and vertical macroblocks width can obtain being respectively 16 and 12.Described reflection method comprises two steps, at first extract in original image is that 16 * 12 sizes are 24 * 24 mapping blocks, after then these mapping blocks being carried out dct transform, most of energy all concentrates on low frequency part, therefore only needing low frequency 16 * 16 coefficients behind the intercepting dct transform, promptly is the piece of required 16 * 16 sizes after down-sampled.
Below, present embodiment is elaborated to the piece Split Method, and establishing coded image is JPEG coded image or MPEG I two field picture, and pending DCT piece is 8 * 8 DCT piece.Present embodiment is just different with first embodiment on the down-sampled method in DCT territory, and it is down-sampled to have adopted a kind of Split Method (3052) to carry out the DCT territory.Described Split Method as shown in Figure 7, the DCT piece with any 3 * 38 * 8 in the DCT territory is made as
To each piece X
IjCarry out 8 * 8 DCT inverse transformation, wherein i, j=1,2,3, obtain on the pixel domain corresponding 24 * 24 block of pixels, be made as
Again this block of pixels of 24 * 24 is carried out 24 * 24 dct transform, intercepting 16 * 16 low frequency part wherein then, and each coefficient of this low frequency matrix all be multiply by 2/3; At last this low frequency part of 16 * 16 is carried out the DCT inverse transformation, promptly obtain 16 * 16 block of pixels, export this block of pixels and get final product.
In addition, the down-sampled operation in DCT territory at this place also can be adopted other down-sampled methods in DCT territory, as the DCT territory convolution method of mentioning in many documents, specific descriptions are, when level is m/n with vertical down-sampled rate, at first the DCT piece that obtains is carried out the liter sampling of n:1 from code stream, and then carry out the down-sampled technology of m:1, thereby finish the image drop sampling of n/m arbitrary proportion.But this method operand is very big, requires also high to memory space.
Claims (9)
1. image drop sampling quick decode display packing, this method may further comprise the steps:
(1) input coding original image, the user sets the resolution of output image;
(2) resolution of the output image of setting according to the resolution of original image and user is determined the down-sampled ratio of horizontal direction and the down-sampled ratio of vertical direction;
(3) the coded image code stream of input is carried out decoding length changeable code and inverse quantization operation;
(4) according to the down-sampled ratio of level that obtains in the step (1) and vertical direction, the DCT piece that obtains after step (3) inverse quantization operation is carried out the down-sampled operation in DCT territory;
(5) the DCT piece after down-sampled is carried out the operation of DCT inverse transformation, rebuild low-resolution image to be exported, obtain reconstructed image;
(6) reconstructed image behind the output decoder.
2. image drop sampling quick decode display packing according to claim 1 is characterized in that: the decoding length changeable code operation that code stream is carried out described in the step (3) specifically describes as follows:
If the DCT block size is M * N, the down-sampled ratio of horizontal direction is X
c: Y
c, the down-sampled ratio of vertical direction is X
r: Y
r, wherein M, N, X
c, Y
c, X
rAnd Y
rBe positive integer;
If satisfy M/ (X
c: Y
c), N/ (X
r: Y
r) be integer, be initial point then in starting point, x axle positive direction to the right, in the downward rectangular coordinate system of y axle positive direction, upper left corner decoding line promptly was the point (M/X-1 in the rectangular coordinate system, N/Y-1) and be parallel to the straight line of DCT piece counter-diagonal, decoding length changeable code is decoded to the zone that the upper left corner decoding line of DCT piece and x axle and y axle surround at least;
If do not satisfy M/ (X
c: Y
c), N/ (X
r: Y
r) be integer, decoding length changeable code then carries out the full decoder operation.
3. image drop sampling quick decode display packing according to claim 1 is characterized in that: down-sampled the operating under the different down-sampled ratios in the described DCT of step (4) territory can be adopted different down-sampled methods, and be as follows:
For width of cloth size is the original image of m * n, and establishing the DCT block size is M * N, and the down-sampled ratio of horizontal direction is X
c: Y
c, the down-sampled ratio of vertical direction is X
r: Y
r, wherein M, N, X
c, Y
c, X
rAnd Y
rBe positive integer;
If satisfy M/ (X
c: Y
c), N/ (X
r: Y
r) be integer, then adopt direct intercepting method that the DCT piece is carried out the down-sampled operation in DCT territory;
If do not satisfy M/ (X
c: Y
c), N/ (X
r: Y
r) be integer, then adopt reflection method or piece method for splitting that the DCT piece is carried out the down-sampled operation in DCT territory.
4. image drop sampling quick decode display packing according to claim 3 is characterized in that: described employing directly intercepts method the DCT piece is carried out the down-sampled operation in DCT territory, directly intercepts the upper left corner M/ (X of each DCT piece behind the inverse quantization
c: Y
c) * N/ (X
r: Y
r) zone, then the DCT inverse transformation is carried out in this zone.
5. image drop sampling quick decode display packing according to claim 3 is characterized in that: described employing reflection method carries out the down-sampled operation in DCT territory to the DCT piece, promptly is to extract size by matrixing from original image to be (M * (X
c: Y
c)) * (N * (X
r: Y
r)) the zone, then dct transform is carried out in this zone, intercepting wherein size is the DCT piece zone of the low frequency part of M * N as down-sampled output image afterwards, then the DCT inverse transformation is carried out in this zone.
6. image drop sampling quick decode display packing according to claim 3 is characterized in that: described employing piece method for splitting carries out the down-sampled operation in DCT territory to the DCT piece, at first to X in the original image
c* X
rEach DCT piece of individual DCT piece carries out the DCT inverse transformation of M * N, obtains the big or small (M * X of being in the pixel domain
c) * (N * X
r) block of pixels, this block of pixels is carried out (M * X
c) * (N * X
r) dct transform, then the intercepting wherein size be (M * Y
c) * (N * Y
r) low frequency DCT coefficient region, again the DCT inverse transformation is carried out in this low frequency coefficient zone.
7. image drop sampling quick decode display packing according to claim 3 is characterized in that:
When the down-sampled ratio of level greater than M and be that integral multiple and the vertical down-sampled ratio of M is greater than N and when being the integral multiple of N, at first adopt the DC component in each DCT piece upper left corner of image behind the direct intercepting method intercepting inverse quantization, then the DC component after the intercepting being reconfigured to new size is the DCT piece of M * N, adopts direct intercepting method to carry out the down-sampled operation in DCT territory to this piece again; Carry out filtering before the image output.
8. image drop sampling quick decode display packing according to claim 3 is characterized in that:
When the down-sampled ratio of level greater than M but satisfies is that the integral multiple of M or vertical down-sampled ratio are greater than N but do not satisfy when being the integral multiple of N, at first directly intercept the DC component in each DCT piece upper left corner behind the inverse quantization, then the DC component after the intercepting being reconfigured to new size is the DCT piece of M * N, adopts reflection method or piece Split Method to carry out the down-sampled operation in DCT territory to this piece again.
9. image drop sampling quick decode display unit, it comprises with the lower part:
The coded image code stream input unit that is used for the image input;
The image drop sampling quick decode device that is used for the down-sampled decoding of image space;
Be used for that decoded image shows;
The image drop sampling quick decode device is connected between coded image code stream input unit and the display unit;
Described image drop sampling quick decode device is characterized in that comprising:
Be used for decoding the variable length decoding unit of the DCT piece of process quantification from image;
The DCT piece that is used for quantizing carries out the inverse quantization unit of inverse quantization;
Be used to need to judge whether the determining device 1 of part decoding length changeable code;
Be used to control the controller of decoding length changeable code;
Be used to select the determining device 2 of the down-sampled unit of DCT;
Be used to the directly intercepting method that realizes and carry out the down-sampled down-sampled unit of integer DCT;
Be used to realize the down-sampled unit of DCT of the down-sampled method of other DCT;
The DCT inverse transformation unit that is used for inverse transformation;
The input of described determining device 1 links to each other with the output of described coded image code stream input unit; The output of described determining device 1 links to each other with the input of described variable length decoding unit; The output of described variable length decoding unit links to each other with the input of described inverse quantization unit; The output of described inverse quantization unit links to each other with the input of described determining device 2; The output of described inverse transformation unit links to each other with the input of display unit; Down-sampled unit of described integer DCT and the down-sampled unit of described DCT are connected in parallel between determining device 2 and the inverse transformation unit; Described controller is connected between determining device 1 and the variable length decoding unit.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNA2009100961991A CN101511011A (en) | 2009-02-19 | 2009-02-19 | Display method and device for image drop sampling quick decode |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CNA2009100961991A CN101511011A (en) | 2009-02-19 | 2009-02-19 | Display method and device for image drop sampling quick decode |
Publications (1)
Publication Number | Publication Date |
---|---|
CN101511011A true CN101511011A (en) | 2009-08-19 |
Family
ID=41003241
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CNA2009100961991A Pending CN101511011A (en) | 2009-02-19 | 2009-02-19 | Display method and device for image drop sampling quick decode |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101511011A (en) |
Cited By (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102118616A (en) * | 2011-02-24 | 2011-07-06 | 深圳市同洲电子股份有限公司 | Picture decoding method and picture decoder |
CN106447626A (en) * | 2016-09-07 | 2017-02-22 | 华中科技大学 | Blurred kernel dimension estimation method and system based on deep learning |
CN107172425A (en) * | 2016-03-07 | 2017-09-15 | 百度在线网络技术(北京)有限公司 | Reduced graph generating method, device and terminal device |
CN107203971A (en) * | 2017-07-28 | 2017-09-26 | 彭志勇 | The gradual loading method of sampled images |
WO2021077489A1 (en) * | 2019-10-22 | 2021-04-29 | Tcl华星光电技术有限公司 | Image decompression method and device, and display terminal |
-
2009
- 2009-02-19 CN CNA2009100961991A patent/CN101511011A/en active Pending
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102118616A (en) * | 2011-02-24 | 2011-07-06 | 深圳市同洲电子股份有限公司 | Picture decoding method and picture decoder |
CN107172425A (en) * | 2016-03-07 | 2017-09-15 | 百度在线网络技术(北京)有限公司 | Reduced graph generating method, device and terminal device |
CN107172425B (en) * | 2016-03-07 | 2020-01-31 | 百度在线网络技术(北京)有限公司 | Thumbnail generation method and device and terminal equipment |
CN106447626A (en) * | 2016-09-07 | 2017-02-22 | 华中科技大学 | Blurred kernel dimension estimation method and system based on deep learning |
WO2018045602A1 (en) * | 2016-09-07 | 2018-03-15 | 华中科技大学 | Blur kernel size estimation method and system based on deep learning |
CN106447626B (en) * | 2016-09-07 | 2019-06-07 | 华中科技大学 | A kind of fuzzy core size estimation method and system based on deep learning |
CN107203971A (en) * | 2017-07-28 | 2017-09-26 | 彭志勇 | The gradual loading method of sampled images |
WO2021077489A1 (en) * | 2019-10-22 | 2021-04-29 | Tcl华星光电技术有限公司 | Image decompression method and device, and display terminal |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN104885467B (en) | Content-adaptive parameter transformation for next-generation Video coding | |
KR101550166B1 (en) | Computational complexity and precision control in transform-based digital media codec | |
US7876976B2 (en) | Method and apparatus for resizing images using discrete cosine transform | |
US8595281B2 (en) | Transforms with common factors | |
JP4361987B2 (en) | Method and apparatus for resizing an image frame including field mode encoding | |
CN105210374B (en) | The control of super-resolution is improved to frequency using characteristics of image | |
JP2004531969A (en) | Method and apparatus for downscaling video image resolution | |
CN107667532A (en) | Conversion for non-baseband signal code selects | |
CN101511011A (en) | Display method and device for image drop sampling quick decode | |
EP2509315B1 (en) | Video decoding switchable between two modes of inverse motion compensation | |
CN100409693C (en) | Orthogonal transformation method for image and video compression | |
Zhu et al. | Compression-dependent transform-domain downward conversion for block-based image coding | |
KR20230145063A (en) | Upsampling of reference pixels for intra-prediction in video coding | |
EP1751984B1 (en) | Device for producing progressive frames from interlaced encoded frames | |
US20130129326A1 (en) | Video player | |
WO2014163903A1 (en) | Integrated spatial downsampling of video data | |
Fu et al. | Mathematical Relation between APBT-Based and DCT-Based JPEG Image Compression Schemes. | |
Kim et al. | Efficient transform domain transcoding: intra frame of H. 264/AVC to JPEG | |
CN112449187B (en) | Video decoding method, video encoding device, video encoding medium, and electronic apparatus | |
Topiwala et al. | FastVDO enhancements of the AV1 codec and comparison to HEVC and JVET codecs | |
KR101423083B1 (en) | Inverse transform unit design method for HEVC decoder | |
Suzuki et al. | M-channel fast Hartley transform based integer DCT for lossy-to-lossless image coding | |
Mietens et al. | New DCT computation algorithm for video quality scaling | |
WO2009047684A2 (en) | Video decoding | |
JP2001103313A (en) | Image processor, image processing method and recording medium |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C02 | Deemed withdrawal of patent application after publication (patent law 2001) | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20090819 |