CN102577344A - Image processing system and image processing program - Google Patents
Image processing system and image processing program Download PDFInfo
- Publication number
- CN102577344A CN102577344A CN2009801618225A CN200980161822A CN102577344A CN 102577344 A CN102577344 A CN 102577344A CN 2009801618225 A CN2009801618225 A CN 2009801618225A CN 200980161822 A CN200980161822 A CN 200980161822A CN 102577344 A CN102577344 A CN 102577344A
- Authority
- CN
- China
- Prior art keywords
- gradient
- pixel value
- image processing
- level
- piece
- 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.)
- Granted
Links
- 238000012545 processing Methods 0.000 title claims abstract description 80
- 238000012937 correction Methods 0.000 claims abstract description 25
- 230000009466 transformation Effects 0.000 claims description 47
- 230000002427 irreversible effect Effects 0.000 claims description 35
- 238000007906 compression Methods 0.000 claims description 10
- 230000006835 compression Effects 0.000 claims description 10
- 238000003672 processing method Methods 0.000 claims description 6
- 238000004364 calculation method Methods 0.000 abstract description 3
- 230000008569 process Effects 0.000 description 26
- 238000000034 method Methods 0.000 description 24
- 101150079125 DCN1 gene Proteins 0.000 description 14
- 230000003139 buffering effect Effects 0.000 description 8
- 238000006073 displacement reaction Methods 0.000 description 8
- 230000007423 decrease Effects 0.000 description 6
- 230000004048 modification Effects 0.000 description 6
- 238000012986 modification Methods 0.000 description 6
- 238000010606 normalization Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 101100428617 Homo sapiens VMP1 gene Proteins 0.000 description 2
- 101150074162 TDC1 gene Proteins 0.000 description 2
- 101150010135 TDC2 gene Proteins 0.000 description 2
- 102100038001 Vacuole membrane protein 1 Human genes 0.000 description 2
- 238000010586 diagram Methods 0.000 description 2
- 238000011002 quantification Methods 0.000 description 2
- 241001436679 Adama Species 0.000 description 1
- 230000033228 biological regulation Effects 0.000 description 1
- 238000013144 data compression Methods 0.000 description 1
- 230000001934 delay Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000002349 favourable effect Effects 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- 230000005764 inhibitory process Effects 0.000 description 1
- 238000013139 quantization Methods 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000000844 transformation Methods 0.000 description 1
- 238000011282 treatment Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
- H04N19/17—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object
- H04N19/176—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object the region being a block, e.g. a macroblock
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N1/00—Scanning, transmission or reproduction of documents or the like, e.g. facsimile transmission; Details thereof
- H04N1/41—Bandwidth or redundancy reduction
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
- H04N19/117—Filters, e.g. for pre-processing or post-processing
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
- H04N19/136—Incoming video signal characteristics or properties
- H04N19/14—Coding unit complexity, e.g. amount of activity or edge presence estimation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/50—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding
- H04N19/593—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using predictive coding involving spatial prediction techniques
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/90—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using coding techniques not provided for in groups H04N19/10-H04N19/85, e.g. fractals
- H04N19/91—Entropy coding, e.g. variable length coding [VLC] or arithmetic coding
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
- Compression Of Band Width Or Redundancy In Fax (AREA)
- Image Processing (AREA)
- Facsimile Image Signal Circuits (AREA)
- Compression, Expansion, Code Conversion, And Decoders (AREA)
Abstract
Disclosed are an image processing system and an image processing program with which overshoot and undershoot due to alternating-current component prediction are controlled and ringing noise is effectually reduced. A slope calculation unit (32a) calculates a slope (a) in a one-dimensional direction. A threshold value setting unit (32b) selects the smaller of the amounts (|L-C|, |R-C|) of change of a pixel value in the lateral direction and sets a threshold slope value (ath) based on the selected smaller amount of change. When the slope (a) in a one-dimensional direction is larger than the threshold value (ath), a slope correction unit (32c) corrects the slope (a) such that the value of the slope (a) is reduced, and then outputs the result. A prediction unit uses the slope (a') in a one-dimensional direction that is output to calculate respective predicted pixel values (C) for multiple sub blocks of a block that has been segmented.
Description
Technical field
The present invention relates to image processing system and image processing program, relate in particular to the noise of ring certainly (ringing noise) countermeasure of alternating current component prediction (ACP:AC Component Prediction).
Background technology
In the past, the known image processing method that the alternating current component of being known as prediction (ACP) is arranged.Alternating current component prediction is meant, as Fig. 1 institute illustration, neighboring area PBl, the information L of PBr, R that reference becomes the object piece PBs of process object obtains the sub-piece sb0 that this piece PBs has been cut apart, the information c0 of sb1, the method for c1.In the alternating current component prediction, simultaneously utilize the correlation in object piece PBs and near the neighboring area PBl it, the space between the PBr to keep the DC value S of object piece PBs, AC value c0, the c1 that the computer memory resolution capability is higher.Disclose in the patent documentation 1～3 the recurrence alternating current component predictive coding (RACP) of level (classification) coding has been cut apart, view data is carried out to the object piece successively.Particularly, the object piece is shifted successively at first on one side at predetermined direction, Yi Bian calculate the predicted pixel values of the sub-piece that the object piece has been cut apart successively.Then, the calculating predicted pixel values is used as prediction residual with the difference of pixel value (true value) originally.Then, through this prediction residual is implemented irreversible transformation and entropy coding, become the alternating current component of image of the part of packed data.More than such processing in the hierarchical structure that for example constitutes, recursively carry out repeatedly by the piece (lowest hierarchical level) of the piece (the 3rd level) of the piece (the 2nd level) of the piece (highest level) of 8 * 8 pixels, 4 * 4 pixels, 2 * 2 pixels and 1 * 1 pixel.
Patent documentation 1: No. 4000157 communique of Japan Patent
Patent documentation 2: No. 3774201 communique of Japan Patent
Patent documentation 3: No. 3700976 communique of Japan Patent
Summary of the invention
In the prediction of in the past alternating current component, around the edge of the profile of image or literal etc., produce sometimes from the noise that rings promptly along the noise of the striated at edge.As Fig. 1 illustration, the noise that rings certainly is difficult to from visually differentiating under for the situation of flat (flat) in monotone variation takes place for the pixel value L that arranges along the one dimension direction, C, R situation (the dull minimizing or dull increasing) or they.But as Fig. 2 illustration,, pixel value S very big in the difference of pixel value L, S, R equate that around the edge of (the small situation of difference that also comprises S, R), as the intrinsic characteristic of alternating current component prediction, horizontal gradient α becomes (sharply) greatly.Its result, predicted pixel values c0 go up to dash+amount of α, and predicted pixel values c1 dashes down-amount of α, and the magnitude relationship of the pixel value of the one dimension direction before the alternating current component prediction is destroyed.This phenomenon causes the decline of image quality except becoming visual noise, also can predicated error in the coding that adopts the alternating current component prediction increases and causes the decline of compression ratio.
Therefore, the objective of the invention is to, the Shang Chong that suppresses to be caused by the alternating current component prediction reduces the noise that rings certainly effectively with bringing down.
In order to solve above-mentioned problem, first invention provides a kind of image processing system, and it has gradient calculating portion, threshold setting portion, gradient correction portion, prediction section, carries out the alternating current component prediction of piece.Gradient calculating portion is based on the information of the conduct reference area relevant with piece and a plurality of pixel values that are determined come the gradient of variable quantity of pixel value of the per unit distance of represents one dimension direction.These pixel values on the plane of delineation are center and relatively with the piece.Threshold setting portion select to be arranged in the one dimension direction a direction pixel value and piece pixel value variable quantity and be positioned at the less side of variable quantity of pixel value of pixel value and piece of another direction of one dimension direction, set the gradient threshold value based on the less side's of this selection variable quantity.Gradient correction portion the gradient of one dimension direction during greater than the gradient threshold value so that the gradient of the mode correction one dimension direction that the value of the gradient of one dimension direction diminishes and exporting.Prediction section use the gradient of the one dimension direction exported calculate with piece cut apart and the predicted pixel values separately of many sub-block.
At this, in first invention, preferably, the gradient of one dimension direction is horizontal gradient and at least one side in the gradient longitudinally on the plane of delineation.Therefore, both can to horizontal gradient and longitudinally these both sides of gradient carry out above-mentioned gradient correction, also can append and in addition the relevant above-mentioned gradient correction of incline direction.
In first invention, preferably, gradient correction portion is modified to the gradient of one dimension direction below the gradient threshold value during greater than the gradient threshold value in the gradient of one dimension direction and exports.
In first invention, adder can also be set, its prediction residual and predicted pixel values of being calculated by prediction section through the predicted pixel values that will be equivalent to calculated by prediction section and the difference of genuine pixel value Calais is mutually calculated the pixel value of sub-piece.
First invention can also be applied to follow the level step-down and hierarchical structure that the big or small stage of piece diminishes.The pixel value of the sub-piece that in higher level's level, calculates in this case, is fed into subordinate's level as the pixel value of the piece of subordinate's level of the next stage that is arranged in higher level's level.
First invention can also be applied to the encoder of compressed image.This encoder also comprises: calculate the subtracter that the predicted pixel values that calculated by prediction section and the difference of genuine pixel value are used as prediction residual; The prediction residual that is calculated by subtracter is implemented the irreversible transformation portion of irreversible transformation; Generate entropy coding portion through the prediction residual of having implemented irreversible transformation being implemented entropy coding as the alternating current component of the image of the part of packed data; And through the prediction residual of having implemented irreversible transformation being carried out the contrary inverse transformation portion that generates the prediction residual that should supply to adder that handles of irreversible transformation.
First invention can also be applied to the decoder of expansion (extension) image.This decoder also comprises inverse transformation portion, and it is handled by irreversible transformation and the contrary of entropy coding that the packed data to image carries out being implemented when the compression of images, and the prediction residual that should supply to adder is restored (post-equalization).
Second invention provides a kind of image processing program that carries out the alternating current component prediction of piece.The image processing method that this image processing program is carried out computer to comprise the steps: based on the information that is confirmed as the reference area relevant with piece, on the plane of delineation be center and relative a plurality of pixel values with the piece; Calculate the step of gradient, gradient is represented the variable quantity of the pixel value of the per unit distance on the one dimension direction; Selection be arranged in the one dimension direction a direction pixel value and piece pixel value variable quantity and be positioned at the less side of variable quantity of pixel value of pixel value and piece of another direction of one dimension direction, set the step of gradient threshold value based on the less side's of this selection variable quantity; In the gradient of one dimension direction during, so that the gradient of the mode correction one dimension direction that the value of the gradient of one dimension direction diminishes and the step exported greater than the gradient threshold value; And the gradient of using the one dimension direction of output calculate with piece cut apart and the step of many sub-block predicted pixel values separately.
At this, in second invention, preferably, the gradient of one dimension direction and first is invented likewise, is horizontal gradient and at least one side in the gradient longitudinally on the plane of delineation.
In second invention, preferably, revise the gradient of above-mentioned one dimension direction and the step exported to be gradient in the one dimension direction be modified to the gradient of one dimension direction below the gradient threshold value during greater than the gradient threshold value and with the step of its output.
In second invention, the predicted pixel values that said prediction residual is equivalent to calculate and the genuine difference of pixel value can be set also through prediction residual and the predicted pixel values that calculates are calculated the step of the pixel value of sub-piece in the Calais mutually.
Second invention can also be applied to follow the level step-down and hierarchical structure that the big or small stage of piece diminishes.In this case, recursively carry out above-mentioned image processing method, the pixel value of the sub-piece that in higher level's level, calculates is fed into subordinate's level as the pixel value of the piece of subordinate's level of the next stage that is arranged in higher level's level.
Second invention can also be applied to the coded program of compressed image.This coded program also comprises: the difference of predicted pixel values that calculates and genuine pixel value is used as the step of prediction residual; The prediction residual that calculates is implemented the step of irreversible transformation; Through the prediction residual of having implemented irreversible transformation is implemented entropy coding, generation is as the step of the alternating current component of the image of the part of packed data; And through the prediction residual of having implemented irreversible transformation being carried out the contrary processing of irreversible transformation, generation should supply to the prediction residual of the 6th step.
Second invention can also be applied to the decoding program of expanded images.This decoding program also comprises contrary processing of the irreversible transformation implemented when carrying out image compression through the packed data to image and entropy coding and step that prediction residual is restored.
According to first invention and second invention, around the edge of gradient greater than the gradient threshold value of one dimension direction, the gradient of correction one dimension direction makes the value of gradient of one dimension direction diminish.Shang Chong that causes and the direction generation effect of dashing are down predicted in the inhibition that is suppressed at of the gradient of one dimension direction by alternating current component, therefore help the reduction from the noise that rings.
Description of drawings
Fig. 1 is the performance plot of the horizontal pixel value in the zone beyond around the edge.
Fig. 2 is the performance plot of the horizontal pixel value around the edge.
Fig. 3 is the overall structure figure of RACP encoder.
Fig. 4 is the key diagram of the piece that on the plane of delineation, sets.
Fig. 5 is the working timing figure of parallel processing.
Fig. 6 is the working timing figure of progressively handling.
Fig. 7 is the structure chart of the level handling part in the RACP encoder.
Fig. 8 is the structure chart of alternating current component prediction section.
Fig. 9 is that expression is the figure of distance relation of the reference block at center with the object piece.
Figure 10 is the calculation specifications figure of horizontal gradient α.
Figure 11 is the input-output characteristic figure as the horizontal gradient α of an example.
Figure 12 is the input-output characteristic figure as the horizontal gradient α of another example.
Figure 13 is the performance plot that go up to dash/dash down the horizontal pixel value that has been suppressed.
Figure 14 is the figure of another setting example of expression reference block.
Figure 15 is the another figure that sets example of expression reference block.
Figure 16 is the flow chart of RACP coded program.
Figure 17 is the overall structure figure of RACP decoder.
Figure 18 is the structure chart of the level handling part in the RACP decoder.
Figure 19 is the flow chart of RACP decoding program.
Embodiment
Below, as the execution mode of having used the alternating current component prediction that the present invention relates to, be that example describes with recurrence alternating current component predictive coding (RACP:Recursive ACP).
(RACP encoder)
Fig. 3 is the overall structure figure of RACP encoder.This encoder has DC calculating part 1, DC encoding section 2, level handling part 3a～3c, as an example, is made up of 4 levels.From these unit 1,2, the data TDCn of 3a～3c output, DCn (n=0,1,2) temporarily be stored in the not shown buffer storage (storage part) at any time.
The input picture that DC calculating part 1 will become compressed object is divided into the piece (block) of predefined size, and the mean value that the average pixel value of each piece promptly is included in the pixel in the piece is exported as TDC0～TDC3.At this, TDC0 is the average pixel value of the piece (below be called " 8 * 8 ") of 8 * 8 pixels, and TDC1 is the average pixel value of the piece (below be called " 4 * 4 ") of 4 * 4 pixels, and TDC2 is the average pixel value of the piece (below be called " 2 * 2 ") of 2 * 2 pixels.In addition, TDC3 be the average pixel value of the piece (below be called " 1 * 1 ") of 1 * 1 pixel, promptly as the pixel value of the pixel of the least unit of image itself.In addition, what note a bit is that the average pixel value TDC0～TDC3 (true value) that directly calculates according to input picture is different from the average pixel value DC0～DC2 (disengaging value) that is restored through the processing in unit 2,3a～3c.Both values are attended by irreversible transformation in the process of coding, therefore inconsistent strictly speaking.
2 couples of per 8 * 8 average pixel value TDC0 that go out from the buffering memory read of the DC encoding section of highest level implement differential pulse coding modulation (DPCM:Differential Pulse Code Modulation) and entropy coding.Utilize the differential pulse coding modulation, the difference of the average pixel value TDC0 relevant with piece adjacent one another are on the plane of delineation is encoded.Then, the difference behind this coding is implemented huffman coding, the such entropy coding of arithmetic coding after quantification.Exported as the DC component DC0 of the image of a part that becomes packed data through the data handled like this, and exported as the value DC0 that 8 * 8 average pixel value is restored after (re-quantization) for the level handling part 3a that also supplies to next level.
3 level handling part 3a～3c are through comprising the average pixel value TDCn (n=1,2,3 that is generated by DC calculating part 1; Below identical) and the average pixel value DCn-1 that generated by the last layer level as the processing of the alternating current component prediction of input, output is as the alternating current component ACn of the image of the part of packed data.And level handling part 3a, 3b (except the 3c) restore average pixel value DCn according to alternating current component ACn.The average pixel value DC0,1,2 that restores is exported in order to supply to next level.
Particularly, the level handling part 3a of second level exports 4 * 4 alternating current component AC1 and 4 * 4 average pixel value DC1 through with 8 * 8 such processing of calculating as prediction of the alternating current components of process object and prediction residual.In this level, use should reference in as the alternating current component prediction at 8 * 8 average pixel value DC0 that highest level generated information.And, in order to calculate prediction residual, 4 * 4 the average pixel value TDC1 that output is calculated by DC calculating part 1.The level handling part 3b of the 3rd level exports 2 * 2 alternating current component AC2 and 2 * 2 average pixel value DC2 through with 4 * 4 processing as the alternating current component prediction of process object etc.In this level, use 4 * 4 average pixel value DC1 being generated in second level as the alternating current component prediction with reference to information.And,, be transfused to 2 * 2 the average pixel value TDC2 that calculates by DC calculating part 1 in order to calculate prediction residual.The level handling part 3c of lowest hierarchical level exports 1 * 1 alternating current component AC3 (need not calculate DC3) through with 2 * 2 processing as the alternating current component prediction of process object etc.In this level, use 2 * 2 average pixel value DC2 being generated in the 3rd level as the alternating current component prediction with reference to information.And,, use 1 * 1 the average pixel value TDC3 that calculates by DC calculating part 1 in order to calculate prediction residual.
Like this, in the hierarchical structure that varies in size of the piece that becomes process object, connect each other with level handling part 3a～3c, recursively carry out the image processing that is predicted as main body with alternating current component through DC encoding section 2.Thus, the DC component DC0 of output image and its alternating current component AC1～AC3 are as packed data.In packed data, except also comprising incidental informations such as huffman table these.
Fig. 4 is the key diagram of the piece that on the plane of delineation, sets.Piece PB has been set a plurality of on the plane of delineation through the image (the for example image of 1 frame or its parts of images) that is partitioned into compressed object in length and breadth, be that unit is processed with the piece.The size of piece PB is configured to diminish along with the level step-down interimly.The sub-piece sb's of having cut apart in the size of piece PB in certain level and the level of one-level on it on the throne is big or small consistent.The piece PB's of the size of the sub-piece sb of last layer level and the low-level that is positioned at its next stage is big or small consistent conversely speaking.The processing of the integral image in certain level through while make become the object of process object piece PBs on picture successively the displacement handle repeatedly, all pieces in the image handled reached.This direction of displacement (scanning direction) can scan by the line along horizontal direction as illustrating successively, but also can comprise along the direction of displacement of vertical direction and at random setting.In addition, the direction of displacement in each level may not need identical.Under the situation of illustrated direction of displacement, the horizontal upside under the object piece PBs becomes processing region, and its downside becomes untreated areas.In addition, the horizontal left side of object piece PBs becomes processing region, and its right side becomes untreated areas.Use 4 identical piece PBt, PBl, PBb, PBr of block size to be used as the reference area of reference in the processing of object piece PBs.About piece PBl, PBr be the center laterally relatively with object piece PBs, piece PBt up and down, PBb are the center vertically relatively with object piece PBs.The average pixel value of the sub-piece sb00 that is divided into 4 parts, sb01, sb10, sb11 is calculated in alternating current component prediction through object piece PBs.Below, when certain piece PB is cut apart, discern upper left sub-piece sb with subscript " 00 ", discern top-right sub-piece with subscript " 01 ", discern the sub-piece of lower left with subscript " 10 ", discern bottom-right sub-piece with subscript " 11 ".
The processing of RACP encoder can be carried out through parallel processing or sequential processes.Fig. 5 is the working timing figure of parallel processing.At first, DC calculating part 1 carries out work, generates average pixel value TDC0～TDC3.Then, when the generation of whole average pixel value TDC0～TDC3 finished, DC encoding section 2 was worked with level handling part 3a～3c concurrently.But the timing that their work begins is different, and level is low more, and beginning is regularly slow more.This delay is that the processing delay that causes of displacement successively owing to higher level's level produces.In certain level, begin to carry out under the situation of processing of object piece PBs, in the last layer level, ended up being condition (processing start condition) with the processing of its reference block PBt～PBr.Therefore, as long as in higher level's level, be shifted, do not finish the processing till the piece PBb successively, just can not begin to carry out the processing of the piece PBs in subordinate's level.This is the reason that produces the delays in work of subordinate's level.But, to compare with sequential processes as shown in Figure 6, it is at a high speed that pipeline parallel method shown in Figure 5 is handled.And in the sequential processes of Fig. 6, the time point in that the processing of certain level is all over begins to carry out the processing of next stage level.Therefore, in sequential processes, also satisfy above-mentioned processing start condition certainly.
Fig. 7 is the structure chart of the level handling part 3 (general name of 3a～3c) in the RACP encoder.Except the different this point of handled block size, the structure of the level handling part 3 in each level and the flow process of processing are basic identical.But, for first degree level handling part 3c, owing to do not exist than its low level, so do not need the generation system 36,37 of DC3.Level handling part 3 comprises that pixel value reads portion 31, alternating current component prediction section 32, subtracter 33, irreversible transformation portion 34, entropy coding portion 35, inverse transformation portion 36 and adder 37.
Pixel value is read portion 31 and is specified to object piece PBs and the reference block PBt～PBr that becomes the zone that it should reference for this process object, from the buffering memory, reads their average pixel value DCn-1.In this execution mode, the information of reference block PBt～PBr and this piece handle/untreated difference is irrelevant, and be in other words irrelevant with the information that whether calculates the sub-piece in this piece, uses the average pixel value DCn-1 that in the upper level level, has calculated without exception.As after Figure 14 of stating and Figure 15 also with reference under the situation of sub-piece, from the buffering memory, read out in the average pixel value DCn that has calculated in the same level.The average pixel value DCn-1 that from the buffering memory, reads out is fed into alternating current component prediction section 32.
The alternating current component that alternating current component prediction section 32 uses the average pixel value DCn-1 (following " T ", " L ", " B ", " R " of being called respectively) of average pixel value DCn-1 (below be called " S ") and the reference block PBt～PBr of object piece PBs to carry out object piece PBs is predicted.Thus, calculate the predicted pixel values C={c00 of 4 sub-block sb00～sb11 that object piece PBs has been cut apart, c01, c10, c11}.
Fig. 8 is the structure chart of alternating current component prediction section 32.Alternating current component prediction section 32 comprises the gradient calculating 32a of portion, the 32b of threshold setting portion, the gradient correction 32c of portion, prediction section 32d.The gradient that the gradient calculating 32a of portion calculates the one dimension direction based on the information of the reference area of object piece PBs is the variable quantity of the pixel value of the per unit distance on the one dimension direction.In this execution mode, calculate horizontal gradient α and these both sides of gradient β longitudinally, as the gradient of one dimension direction, but also can be any side, perhaps also can append incline direction etc. to them.Horizontal gradient α is based on being center and pixel value L, R about laterally relative calculates with object piece PBs on the plane of delineation.In addition, longitudinally gradient β based on being center and calculating with object piece PBs on the plane of delineation at vertically relative up and down pixel value T, B.The calculating formula of these gradients α, β is expressed as formula 1.
(formula 1)
α＝(L-R)/8
β＝(T-B)/8
Fig. 9 is that expression is the figure of distance relation of the reference block PBt～PBr at center with object piece PBs.The center pixel value of the center position of each piece PB is regarded the average pixel value DCn-1 of this piece itself as.The length on one side of the square in the time of will being divided into 4 parts to piece PB (sub-piece sb) is made as 1, and the length on the foursquare one side when bundle piece sb further is divided into 4 parts is made as 1/2.Adjacent piece distance between centers each other is 2, and the distance between centers of piece and its inner sub-piece is 1/2 (normalization coefficient).Divisor 8 in the above-mentioned formula 1 be the distance between centers 4 with the center pixel value L of about 1/2 pair of normalization coefficient, R carry out standardization and value.
Figure 10 is the calculation specifications figure of horizontal gradient α.At first, with about the difference of center pixel value L, R of reference block PBl, PBr divided by both distance between centers 4.Then, the value that obtains is thus carried out standardization with normalization coefficient 1/2, thereby calculate horizontal gradient α.Gradient β with the difference of the center pixel value T of up and down reference block PBt, PBb, the B distance between centers 4 divided by both, carries out standardization with 1/2 pair of this division value of normalization coefficient too longitudinally.The gradient α, the β that are calculated by the gradient calculating 32a of portion are outputed to the gradient correction 32c of portion.
The 32b of threshold setting portion calculates gradient threshold alpha th, β th to gradient α, β respectively.About horizontal gradient α; Variable quantity between the pixel value S of the pixel value L in the left side of calculating object piece PBs and object piece PBs itself | L-S|, and the variable quantity between the pixel value S of the pixel value R on the right side of calculating object piece PBs and object piece PBs itself | R-S|.Then, select these variable quantities | L-S|, | a less side among the R-S| is set and is horizontal gradient threshold alpha th.On the other hand; About gradient β longitudinally; Variable quantity between the pixel value S of the pixel value T of the upside of calculating object piece PBs and object piece PBs itself | T-S|, and the variable quantity between the pixel value S of the pixel value B of the downside of calculating object piece PBs and object piece PBs itself | B-S|.Then, select these variable quantities | T-S|, | a less side among the B-S| is set and is gradient threshold value beta th longitudinally.The gradient threshold alpha th, the β th that are set by the 32b of threshold setting portion are outputed to the gradient correction 32c of portion.
The gradient correction 32c of portion is that gradient threshold alpha th does not revise horizontal gradient α when following and it is directly outputed to prediction section 32d (input value α=output valve α ') at horizontal gradient α, makes the value of this gradient α diminish and exports (input value α＞output valve α ') and revise gradient α during greater than gradient threshold alpha th at horizontal gradient α.Likewise; At gradient β longitudinally is that gradient threshold value beta th is when following; Directly output of gradient β quilt longitudinally (input value | β |=output valve | β ' |); When gradient β is greater than gradient threshold value beta th longitudinally, revise gradient β make the value of this gradient β diminish and export (input value | β |＞output valve | β ' |).
In this execution mode,, revise gradient α, β based on following formula 2 as an example.At this, sign (x) is meant the sign function of the symbol of constant inclination gradient, o'clock returns output 1 in x＞0, returns output 0 during at x=0, o'clock returns in x＜0 output-1.
(formula 2)
α’＝sign(α)×αth
β’＝sigin(β)×βth
Figure 11 is the input-output characteristic figure of the horizontal gradient α among the gradient correction 32c of portion.| α | in the non-modification region of≤α th, do not dash (undershoot) owing to do not produce upward dash (overshoot)/down that cause by the alternating current component prediction, so with input value | α | directly as output valve | α ' |.With respect to this, at | α | in the modification region of＞α th, with input value | α | value irrespectively, with output valve | α ' | be restricted to gradient threshold alpha th (constant).Gradient | α | th means greater than the gradient threshold alpha, because the gradient of this direction is big, so if keeping this state to produce dashes/dash down.In this case, if with respect to input value | α | and reduce output valve | α ' |, poor corresponding to output valve and original value α then, go up dash/dash down and relaxed.Particularly, if with output valve | α ' | be limited in below the gradient threshold alpha th, then can eliminate dashing in theory fully/dash down.Also can directly use gradient threshold alpha th as correction value, can also use it multiply by predetermined coefficient n (0＜n≤1) and the value that obtains.In addition, can be not the output valve α ' of modification region be made as steady state value, and that kind shown in figure 12 yet, Yi Bian keep | α |＞| α ' | condition, Yi Bian along with input value | α | increase and make output valve | α ' | curved shape increases.When setting input-output characteristic, need bottom line to satisfy output valve | α ' | less than input value | and α | such condition, as long as satisfy this condition, can set any input-output characteristic.And then the correction of gradient α can also be carried out with reference to the table of recording and narrating the corresponding relation that input value α and output valve α ' arranged except the use arithmetic expression is carried out.More than these, for gradient β longitudinally too.The gradient α ', the β ' that are calculated by the gradient correction 32c of portion are outputed to prediction section 32d.
(formula 3)
c00＝T+(α’+β’)
c01＝T+(-α’+β’)
c10＝T+(α’-β’)
c11＝T+(-α’-β’)
(formula 4)
c00＝T+(L-R+T-B)/8
c01＝T+(-L+R+T-B)/8
c10＝T+(L-R-T+B)/8
c11＝T+(-L+R-T+B)/8
Once more with reference to Fig. 7; Subtracter 33 calculates the difference of the genuine pixel value that the predicted pixel values C (for example 4 * 4 block units) that calculated by alternating current component prediction section 32 and the average pixel value TDCn (for example 4 * 4 block units) corresponding with it promptly calculate based on input picture, as prediction residual PR.Through the more approaching original image of predicted picture (set of predicted pixel values c00～c11) (genuine pixel value) that the alternating current component prediction generates, prediction residual PR is more little.Through the gradient α among the gradient correction 32c of portion, the correction of β, to eliminating the direction effect that prediction departs from, therefore compare with the situation of direct use gradient α, β, prediction residual PR diminishes.Prediction residual PR is more little, and near the tendency that the frequency of occurrences of the prediction residual of finding out is partial to 0 with adding up is high more, is therefore becoming favourable (helping the raising of compression ratio) aspect the enforcement entropy coding.
The 34 pairs of irreversible transformations of prediction residual PR embodiment as Hadamard transform and quantification that calculate by subtracter 33 of irreversible transformation portion.Entropy coding portion 35 is through generating for example the alternating current component ACn of 4 * 4 block units (packed data) to prediction residual PR enforcement huffman coding, the such entropy coding of arithmetic coding of having implemented irreversible transformation.Prediction residual PR ' calculates through contrary processing of the prediction residual PR that has implemented irreversible transformation being implemented this irreversible transformation in inverse transformation portion 36.This prediction residual PR ' is the value after original prediction residual PR is restored, but owing to implemented irreversible transformation (can't original value be restored fully), different slightly with original prediction residual PR.Adder 37 is through calculate sub-piece (for example 4 * 4) average pixel value DCn in the Calais with prediction residual PR ' and the predicted pixel values C that is calculated by alternating current component prediction section 32 mutually.Do not use original prediction residual PR and use the prediction residual PR ' that restored to be, because the repeated treatments during decoding and accumulated error, guarantee the reproducibility of expanded images in order to prevent as the reason of the input of adder 37.Be temporarily stored buffer storage from the average pixel value DCn of adder 37 outputs, in the processing of next level, read at any time.
Like this, the RACP encoder that relates to according to this execution mode is gradient threshold alpha th at gradient α, the β of one dimension direction, when β th is following, and as the processing of non-modification region, the value of directly exporting α, β is used as gradient α ', β '.Then, based on these gradients α ' (=α), β ' (=β) calculate predicted pixel values C.In this case, same for example for horizontal gradient α with illustrative prior art among Fig. 1, just can not cause originally dashing/dashing down, calculate predicted pixel values c0, the c1 (AC value) of S ± α.With respect to this, greater than gradient threshold alpha th, during β th, make the value of α, β diminish at the gradient α of one dimension direction as around the edge, β as the processing correction of modification region, with the value of revising as gradient α ', β ' output.Then, based on these gradients | α ' | (＜| α |), | β ' | (＜| β |) calculates predicted pixel values C.In this case, for example for horizontal gradient α, as illustration among Figure 13, regulation is with respect to the variable quantity of S (DC value) | α ' | be restricted to below the gradient threshold alpha th.Thus, (L＞S >=R) is maintained magnitude relationship after the alternating current component prediction (L＞c0＞c1=R), or the form approaching with it to the magnitude relationship before the alternating current component prediction.Thus, going up of predicted pixel values c0 can be suppressed, generation can be reduced effectively from the noise that rings towards dashing with the following of predicted pixel values c1.Its result can suppress the decline of image quality, and the decline of the compression ratio that can also prevent to be caused by the increase of predicated error.Particularly, as if size with gradient | α |, | β | replace with value or the value below it of gradient threshold alpha th, β th, then can eliminate fully in theory from ringing noise.In addition, handle, can realize whole reduction in processing time if carry out the pipeline parallel method of that kind shown in Figure 5.
In above-mentioned execution mode, as the reference area of object piece PBs, set the piece PBt～PBr up and down with its direct neighbor, but the various settings of the setting example shown in below can being applied to comprise.In addition; In above-mentioned execution mode; In the calculating of the calculating of gradient α, β and gradient threshold alpha th, β th, same pixel value T, L, B, R (with reference to information) have been used; But also for example the calculating of gradient α, β uses the calculating of bulk, gradient threshold alpha th, β th to use fritter, makes use in both calculating different with reference to information like this.
In the setting example of Figure 14,, use identical 8 piece PBt, PBtt, PBl, PBll, PBb, PBbb, PBr, the PBrr of block size as the reference area of object piece PBs.The direct neighbor up and down of reference block PBt, PBl, PBb, PBr and object piece PBs.In addition, reference block PBtt, PBll, PBbb, PBrr and object piece PBs's separates one and adjacent up and down.In this setting example, the calculating of gradient α, β is made as formula 1, only use predicted pixel values T ', L ', B ', R ' (formula 5) for the calculating of gradient threshold alpha th, β th.Predicted pixel values T ', L ', B ', R ' use the pixel value S of object piece PBs and predict at the pixel value (for example L, LL on the left) of existing 2 pieces on this direction.It is identical with above-mentioned execution mode to calculate later processing such as gradient α, β.
(formula 5)
L’＝L+(S-LL)/4
R’＝R+(S-RR)/4
T’＝T+(S-TT)/4
B’＝B+(S-BB)/4
α＝(L-R)/8
β＝(T-B)/8
αth＝min(|L’-S|，|R’-S|)
βth＝min(|T’-S|，|B’-S|)
In the setting example of Figure 15, the piece PBt of top, the piece PBl of left are used center pixel value t0, t1, l0, the l1 of sub-piece sb (T), sb (L).Handle as DCn in the processing of these center pixel value t0～l1 before same level, its value is fed back.In this setting example, use predicted pixel values T ', L ', B ', R ' (formula 6) for these both sides of calculating of the calculating of gradient α, β and gradient threshold alpha th, β th.In this case, prediction is center relative predicted pixel values on the one dimension direction with object piece PBs, uses these values calculating gradients α, β.It is identical with above-mentioned execution mode to calculate later processing such as gradient α, β.
(formula 6)
T’＝(t0+t1)/2
L’＝(l0+l1)/2
B’＝B+(S-BB)/8
R’＝R+(S-RR)/8
α＝(L’-R’)/6
β＝(T’-B’)/6
αth＝min(|L’-S|，|R’-S|)
βth＝min(|T’-S|，|B’-S|)
(RACP coded program)
Then, explain that the software processes that is used for through computer realizes the RACP coded program with the equal processing of the RACP encoder of realizing as hardware.Owing to there is not the difference of essence between hardware handles and the software processes, so in this diagrammatic illustration only, its detailed condition is with reference to above-mentioned record.
Figure 16 is the flow chart of RACP coded program.In the software processes of being undertaken by computer, sequential processes shown in Figure 6 is basic.At first, in step 1, on the plane of delineation of input picture, set a plurality of (4), calculate the average pixel value TDC0～TDC3 of each piece.Average pixel value TDC0～3 that calculate are stored in the buffer storage at any time.When the processing of whole pieces finishes, get into step 2.
In step 2, it is differential pulse width modulated and entropy coding that 8 * 8 average pixel value TDC0 are implemented the DC coding, become thus packed data a part image DC component DC0 and with its output.In addition, its disengaging value is stored in the buffer storage as DC0.
In step 3, level sequence number LN is set as 1 (initial value).LN=1 representes that the level that should handle is a highest level.And when 1 level processing finished, (step 14) was passed subordinate's level in order to increase by 1.Then, be the time point of the processing end of lowest hierarchical level at LN=3, whole processing finishes (step 13).
In step 4, piece sequence number BN is set as 1 (initial value).Accompany with the setting of level sequence number LN in the preceding step 3, confirm in this level, to become the size of the piece of process object uniquely, also confirm and the whole corresponding end block sequence number of piece number BNend.When the processing of the piece in same level finished, piece sequence number BN increases by 1, and (step 12), process object was shifted in predetermined direction successively.Then, at the time point that the processing of the piece that is equivalent to end block sequence number BNend finishes, the processing of this level finishes (step 11).
In step 5, to object piece PBs with piece sequence number BN appointment, read reference block pixel value DCn (=T, L, B, R).
In step 6, to object piece PBs, carry out alternating current component prediction based on above-mentioned formula 1～3 with piece sequence number BN appointment, calculate predicted pixel values C (=c00～c11) thus.Then, in following step 7, the difference of calculating the predicted pixel values C that calculates in the step 6 in front and genuine pixel value TDCn is used as prediction residual PR.
In step 8, the prediction residual PR that calculates in the step 7 is in front implemented irreversible transformation and entropy coding.Thus, generation is as the alternating current component ACn of the image of the part of packed data.In following step 9, the predicated error PR that has implemented irreversible transformation carried out its contrary processing, calculate the predicated error PR ' (disengaging value) that original predicated error PR has been restored.Then, in step 10, through with prediction residual PR ' and the predicted pixel values C addition that in step 6, calculates, calculate the average pixel value DCn of sub-piece, and it is stored in the buffer storage.Step 9,10 processing are skipped during lowest hierarchical level is handled at LN=3.
In step 11, whether decision block sequence number BN arrives end block sequence number BNend.Piece sequence number BN is each in step 12 increases by 1, carry out step 5 repeatedly～11 processing, till arriving end block sequence number BNend.Then, when the processing of the whole pieces when arriving end block sequence number BNend, in the plane of delineation finishes, withdraw from circulation and get into step 13.
In step 13, judge whether level sequence number LN arrives 3 (lowest hierarchical level).Level sequence number LN is each in step 14 increases by 1, carry out step 5 repeatedly～12 circulation, until arriving till 3.Thus, the big or small stage of object piece PBs diminishes along with the level step-down, recursively carries out above-mentioned a series of processing.The average pixel value DCn (output) of the sub-piece that calculates in the high-level side is as the average pixel value DCn-1 (input) of low-level side and use.And, when whole processing of lowest hierarchical level finish when arriving LN=3 and BN=BNend, withdraw from circulation according to step 11,13 judged result, finish the processing of this program thus.
According to the RACP coded program that this execution mode relates to, same with above-mentioned RACP encoder, the decline of the image quality that can suppress to cause, and the decline of the compression ratio that can prevent to cause by the increase of predicated error by the noise that rings certainly.
(RACP decoder)
Figure 17 is the overall structure figure of the RACP decoder expanded of the packed data that generates utilizing above-mentioned RACP encoder or RACP coded program.This decoder is that main body constitutes with DC lsb decoder 5 and 3 level handling part 6a～6c.From these unit 5, the data DCn (n=0,1,2,3) of 6a～6c output temporarily is stored in the not shown buffer storage (storage part).
DC lsb decoder 5 is handled with level handling part 6a～6c is assigned to oneself in the hierarchical structure of 8 * 8 (highest levels), 4 * 4 (second level), 2 * 2 (the 3rd levels), 1 * 1 (lowest hierarchical level) level.The DC lsb decoder 5 of highest level is handled through the contrary of processing that the packed data relevant with the DC component DC0 of image is carried out implemented when the image compression, generates 8 * 8 average pixel value DC0, and it is supplied to the level handling part 6a of second level.
The hierarchical structure of RACP decoder itself and RACP encoder roughly the same, but difference is, becomes the alternating current component AC1～AC3 of the output of the level handling part 3a～3c in the encoder, becomes the input of the level handling part 6a～6c in the decoder.The average pixel value DCn-1 that these level handling parts 6a～6c supplies with based on the packed data relevant with the alternating current component AC1～AC3 of image with from the last layer level and average pixel value DCn is restored.Average pixel value DCn after the recovery is exported in order to supply to next level as required, and is stored in the buffer storage.Then, the set of 1 * 1 the average pixel value DC3 that is calculated by the level handling part 6c of lowest hierarchical level becomes final expanded images.The direction of displacement successively that the direction of displacement successively in the decoder is made as with encoder is the direction of standard.In addition, the processing of decoder can be parallel processing or sequential processes.
Figure 18 is the structure chart of the level handling part 6 (general name of 6a～6c) in the RACP decoder.Except the handled this point that varies in size, the basic structure of the level handling part 6 in each level is identical with work.Level handling part 6 comprises that pixel value reads portion 61, alternating current component prediction section 62, adder 63 and inverse transformation portion 64.
It is same with the RACP encoder that pixel value is read portion 61, from the buffering memory read go out to be present in object piece PBs around reference block PBt～rr average pixel value DCn (=T, L, B, R).This average pixel value DCn-1 calculates in the last layer level, supplies with via buffer storage from the last layer level.The average pixel value DCn-1 that comes out from the buffering memory read is fed into alternating current component prediction section 62.
Alternating current component prediction section 62 has structure shown in Figure 8, uses the DCn-1 that supplies with from the buffering memory to carry out the alternating current component prediction of object piece PBs.Thus, the predicted pixel values C of calculating object piece PBs (=c00～c11).The predicted pixel values C that calculates is fed into adder 63.When the part of neighboring area as the edge part on the plane of delineation shortcoming, according to and decoder between the rule that is predetermined handle.For example, replenish the information of the average pixel value DCn-1 of object piece PBs as this part.
Like this, according to the RACP decoder that this execution mode relates to, can suitably expand the packed data that is generated by above-mentioned RACP decoder or RACP coded program.Particularly, handle, can realize whole reduction in processing time if carry out the pipeline parallel method of that kind shown in Figure 5.
(RACP decoding program)
Then, explain that the software processes that is used for through computer realizes the RACP decoding program with the equal processing of the RACP decoder of realizing as hardware.Owing to there is not the difference of essence between hardware handles and the software processes, so in this diagrammatic illustration only, its details is with reference to above-mentioned record.
Figure 19 is the flow chart of RACP decoding program.In the software processes of being undertaken by computer, sequential processes shown in Figure 6 is basic.At first, in step 21, the packed data of the DC component DC0 of image is implemented the contrary of encoding process that the DC decoding promptly implemented when the data compression handle, 8 * 8 average pixel value DC0 is restored thus, and it is stored in the buffer storage.
In step 22,23, same with the RACP coded program, level sequence number LN is set as 1 (initial value), and piece sequence number BN is set as 1 (initial value).
In step 24, from the buffering memory read go out with the specified object piece PBs of piece sequence number BN with reference to information T, L, B, R.
In step 25, to the specified object piece PBs of piece sequence number BN, carry out the alternating current component prediction based on above-mentioned formula 1～3, the predicted pixel values C that calculating is relevant with object piece PBs thus (=c00～c11).Then, in the step of following 26, the packed data of the alternating current component ACn of expanded images, prediction residual PR ' is restored thus.Then, in step 27,, calculate the average pixel value DCn of sub-piece, it is stored in the buffer storage through prediction residual PR ' that will in step 27, calculate and the predicted pixel values C addition that in step 25, calculates.
In step 28, whether decision block sequence number BN arrives end block sequence number BNend.Piece sequence number BN is each in step 29 increases by 1, carry out step 24 repeatedly～27 processing, till arriving end block sequence number BNend.Then, when arriving end block sequence number BNend, when promptly the processing of the whole pieces in the plane of delineation finishes, withdraw from circulation, get into step 30 according to the judged result of step 28.
In step 30, judge whether level sequence number LN arrives 3 (lowest hierarchical level).Level sequence number LN is each in step 31 increases by 1, carry out step 23 repeatedly～29 circulation, until arriving till 3.Thus, along with the level step-down, the big or small stage of piece PB diminishes, and recursively carries out above-mentioned a series of processing.The average pixel value DCn (output) of the sub-piece that in higher level's level side, calculates uses as the average pixel value DCn-1 (input) of the process object that becomes rudimentary level side and the piece of the neighboring area of being untreated.Then, when arriving LN=3 and BN=BNend, i.e. when the processing of lowest hierarchical level finishes, withdraw from circulation, the processing of this program finishes thus.
RACP decoding program according to this execution mode relates to obtains and the above-mentioned same effect of RACP decoder.
Utilizability on the industry
Image processing system that the present invention relates to and image processing program; Not only can be applied to alternating current component predictive coding (comprising the predictive coding of recurrence alternating current component), also can be widely used in using the alternating current component prediction and generate the image processing etc. of the high image of resolution.
Description of reference numerals
The 1:DC calculating part
The 2:DC encoding section
3 (3a～3c): the level handling part
The 5:DC lsb decoder
6 (6a～6c): the level handling part
31,61: pixel value is read portion
32,62: the alternating current component prediction section
32a: gradient calculating portion
32b: threshold setting portion
32c: gradient correction portion
32d: prediction section
33: subtracter
34: irreversible transformation portion
35: entropy coding portion
36,64: inverse transformation portion
37,63: adder
Claims (14)
1. an image processing system carries out the alternating current component prediction of piece, it is characterized in that, comprising:
Gradient calculating portion; Its based on be confirmed as with the information of said relevant reference area, on the plane of delineation, be center and relative a plurality of pixel values with said; Calculate gradient, said gradient is represented the variable quantity of the pixel value of the per unit distance on the one dimension direction;
Threshold setting portion; Its selection be arranged in said one dimension direction a direction pixel value and said pixel value variable quantity and be positioned at pixel value and the said less side of variable quantity of pixel value of another direction of said one dimension direction, set the gradient threshold value based on the less side's of this selection variable quantity;
Gradient correction portion, it is in the gradient of said one dimension direction during greater than said gradient threshold value, so that the gradient of the said one dimension direction of mode correction that the value of the gradient of said one dimension direction diminishes and exporting; And
Prediction section, it uses the gradient of the one dimension direction of said output to calculate many sub-block predicted pixel values separately that said is cut apart and get.
2. image processing system according to claim 1 is characterized in that,
The gradient of said one dimension direction is horizontal gradient and at least one side in the gradient longitudinally on the plane of delineation.
3. image processing system according to claim 1 and 2 is characterized in that,
Said gradient correction portion in the gradient of said one dimension direction during greater than said gradient threshold value, is modified to the gradient of said one dimension direction below the said gradient threshold value and exports.
4. according to any described image processing system in the claim 1～3, it is characterized in that,
Also comprise adder; Said adder is through with prediction residual and the said predicted pixel values addition that is calculated by said prediction section; Calculate the pixel value of said sub-piece, the difference of the said predicted pixel values that said prediction residual is equivalent to be calculated by said prediction section and genuine pixel value.
5. image processing system according to claim 4 is characterized in that,
In the hierarchical structure that said along with the level step-down big or small stage diminishes, the pixel value of the said sub-piece that in higher level's level, calculates as be arranged in this higher level's level next stage subordinate's level said pixel value and be fed into this subordinate's level.
6. according to claim 4 or 5 described image processing systems, it is characterized in that,
Said image processing system is the encoder of compressed image, also comprises:
Subtracter, its difference of calculating the said predicted pixel values that calculated by said prediction section and genuine pixel value is used as prediction residual;
Irreversible transformation portion, it implements irreversible transformation to the said prediction residual that is calculated by said subtracter;
Entropy coding portion, it is through implementing entropy coding to the prediction residual of having implemented said irreversible transformation, and generation is as the alternating current component of the image of the part of packed data; And
Inverse transformation portion, it is handled through the said prediction residual of having implemented said irreversible transformation being carried out the contrary of this irreversible transformation, and generation should supply to the said prediction residual of said adder.
7. according to claim 4 or 5 described image processing systems, it is characterized in that,
Said image processing system is the decoder of expanded images; Also comprise inverse transformation portion; Said inverse transformation portion handles through irreversible transformation and the contrary of entropy coding that the packed data to image carries out being implemented when the image compression, and the said prediction residual that should supply to said adder is restored.
8. an image processing program carries out the alternating current component prediction of piece, it is characterized in that make computer carries out image processing method, said image processing method comprises:
Based on be confirmed as with the information of said relevant reference area, on the plane of delineation, be center and relative a plurality of pixel values with said; Calculate the step of gradient, said gradient is represented the variable quantity of the pixel value of the per unit distance on the one dimension direction;
Selection be arranged in said one dimension direction a direction pixel value and said pixel value variable quantity and be positioned at pixel value and the said less side of variable quantity of pixel value of another direction of said one dimension direction, set the step of gradient threshold value based on the less side's of this selection variable quantity;
In the gradient of said one dimension direction during, so that the gradient of the said one dimension direction of mode correction that the value of the gradient of said one dimension direction diminishes and the step exported greater than said gradient threshold value; And
Use the gradient of the one dimension direction of said output to calculate step with said many sub-block predicted pixel values separately of cutting apart and getting.
9. image processing program according to claim 8 is characterized in that,
The gradient of said one dimension direction is horizontal gradient and at least one side in the gradient longitudinally on the plane of delineation.
10. according to Claim 8 or 9 described image processing programs, it is characterized in that,
The step of revising the gradient of said one dimension direction and exporting is, in the gradient of said one dimension direction during greater than said gradient threshold value, the gradient of said one dimension direction is modified to below the said gradient threshold value and the step of output.
11. any described image processing program according to Claim 8～10 is characterized in that,
Also comprise: through prediction residual and the said predicted pixel values that calculates are calculated the step of the pixel value of said sub-piece in the Calais mutually, said prediction residual is equivalent to the difference of the said predicted pixel values that calculates and genuine pixel value.
12. image processing program according to claim 11 is characterized in that,
In the hierarchical structure that said along with the level step-down big or small stage diminishes, recursively carry out said image processing method,
The pixel value of the said sub-piece that in higher level's level, calculates is fed into this subordinate's level as said pixel value of subordinate's level of the next stage that is arranged in this higher level's level.
13. according to claim 11 or 12 described image processing programs, it is characterized in that,
Said image processing program is the coded program of compressed image, also comprises:
The difference of calculating said predicted pixel values that calculates and genuine pixel value is used as the step of prediction residual;
The said prediction residual that calculates is implemented the step of irreversible transformation;
Through the prediction residual of having implemented said irreversible transformation is implemented entropy coding, generation is as the step of the alternating current component of the image of the part of packed data; And
Handle through the said prediction residual of having implemented said irreversible transformation being carried out the contrary of this irreversible transformation, generation should supply to the said prediction residual of said the 6th step.
14. according to claim 11 or 12 described image processing programs, it is characterized in that,
Said image processing program is the decoding program of expanded images, also comprises contrary processing of the irreversible transformation that carries out being implemented when the image compression through the packed data to image and entropy coding and step that said prediction residual is restored.
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| PCT/JP2009/067344 WO2011042945A1 (en) | 2009-10-05 | 2009-10-05 | Image processing system and image processing program |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN102577344A true CN102577344A (en) | 2012-07-11 |
| CN102577344B CN102577344B (en) | 2014-09-10 |
Family
ID=43856446
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN200980161822.5A Active CN102577344B (en) | 2009-10-05 | 2009-10-05 | Image processing system and image processing program |
Country Status (5)
| Country | Link |
|---|---|
| JP (1) | JP5007432B2 (en) |
| KR (1) | KR101297862B1 (en) |
| CN (1) | CN102577344B (en) |
| HK (1) | HK1173011A1 (en) |
| WO (1) | WO2011042945A1 (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP7040793B2 (en) * | 2019-10-30 | 2022-03-23 | 株式会社アクセル | Image processing device, image processing method, image processing program, and image processing system |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH02154572A (en) * | 1988-12-06 | 1990-06-13 | Nec Corp | Decoding system for picture signal |
| JP3700976B2 (en) * | 2002-08-14 | 2005-09-28 | 株式会社アクセル | Image processing apparatus and image processing method |
| JP3774201B2 (en) * | 2003-03-19 | 2006-05-10 | 株式会社アクセル | Image processing device |
| WO2006040960A1 (en) * | 2004-10-08 | 2006-04-20 | Matsushita Electric Industrial Co., Ltd. | Image processing apparatus and image processing program |
| JP4399449B2 (en) * | 2006-11-14 | 2010-01-13 | 株式会社アクセル | Image noise reduction system, image noise reduction method, and computer program |
| JP2008283341A (en) * | 2007-05-09 | 2008-11-20 | Matsushita Electric Ind Co Ltd | Noise elimination device and noise elimination method of compressed coded image |
-
2009
- 2009-10-05 JP JP2011535222A patent/JP5007432B2/en active Active
- 2009-10-05 WO PCT/JP2009/067344 patent/WO2011042945A1/en active Application Filing
- 2009-10-05 KR KR1020127008725A patent/KR101297862B1/en active IP Right Grant
- 2009-10-05 CN CN200980161822.5A patent/CN102577344B/en active Active
-
2013
- 2013-01-03 HK HK13100055.7A patent/HK1173011A1/en unknown
Also Published As
| Publication number | Publication date |
|---|---|
| JPWO2011042945A1 (en) | 2013-02-28 |
| CN102577344B (en) | 2014-09-10 |
| KR20120058608A (en) | 2012-06-07 |
| JP5007432B2 (en) | 2012-08-22 |
| KR101297862B1 (en) | 2013-08-19 |
| WO2011042945A1 (en) | 2011-04-14 |
| HK1173011A1 (en) | 2013-05-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US7636451B2 (en) | Digital watermark embedding apparatus and method, and digital watermark detection apparatus and method | |
| CN104219533B (en) | A kind of bi-directional motion estimation method and up-conversion method of video frame rate and system | |
| JP5068265B2 (en) | Select motion vector | |
| US20140334541A1 (en) | Video compression device | |
| US9736490B2 (en) | Apparatus and method of compressing and restoring image using filter information | |
| JP4785678B2 (en) | Image coding apparatus and image coding method | |
| TWI506965B (en) | A coding apparatus, a decoding apparatus, a coding / decoding system, a coding method, and a decoding method | |
| CN102577344A (en) | Image processing system and image processing program | |
| WO2001097510A1 (en) | Image processing system, image processing method, program, and recording medium | |
| JP4510701B2 (en) | Video encoding device | |
| CN103069798A (en) | Methods and devices for encoding and decoding at least one image implementing an estimation in the transform domain, and corresponding signal and computer program | |
| JP2008252225A (en) | Motion picture encoding device and method | |
| US9036934B2 (en) | Image encoder and image processing system | |
| JP4660698B2 (en) | Image processing system and image processing program | |
| JP4973591B2 (en) | Motion vector detection apparatus and motion vector detection method | |
| JP2016208366A (en) | Image decoder | |
| JP2008098796A (en) | Video encoding method, video encoding device, video encoding program, and recording medium thereof | |
| US20040151397A1 (en) | Smoothing processor, smoothing processing method and smoothing processing program of image data | |
| WO2018083183A1 (en) | Intra compression of screen content based on soft context formation | |
| JP2020092327A (en) | Image encoding device, image encoding method, and program | |
| JP3123857B2 (en) | Motion compensation method and apparatus | |
| De | Adjacent Pixel Values Blocking and Prediction Error Expansion Based High Fidelity Reversible Data Hiding Scheme | |
| JP4160513B2 (en) | Moving image luminance change parameter estimation method, moving image luminance change parameter estimation program and recording medium thereof, and moving image encoding apparatus, moving image encoding method, moving image encoding program and recording medium thereof | |
| JP3957123B2 (en) | Shape information encoding method | |
| JP5597028B2 (en) | Image decoding apparatus, control method therefor, program, and computer-readable storage 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 | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: DE Ref document number: 1173011 Country of ref document: HK |
|
| C14 | Grant of patent or utility model | ||
| GR01 | Patent grant | ||
| REG | Reference to a national code |
Ref country code: HK Ref legal event code: GR Ref document number: 1173011 Country of ref document: HK |