CN104243996B - The pixel value difference compensation method of processing unit and selectivity and coded system in loop - Google Patents
The pixel value difference compensation method of processing unit and selectivity and coded system in loop Download PDFInfo
- Publication number
- CN104243996B CN104243996B CN201310236938.9A CN201310236938A CN104243996B CN 104243996 B CN104243996 B CN 104243996B CN 201310236938 A CN201310236938 A CN 201310236938A CN 104243996 B CN104243996 B CN 104243996B
- Authority
- CN
- China
- Prior art keywords
- bit rate
- coding
- value
- offset
- image data
- 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.)
- Active
Links
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The pixel value difference compensation method of processing unit and selectivity and coded system in a kind of loop, view data and reconstructed image data after coding before this method reference encoder, multiple offsets are produced according to pixel difference state in the square of two view data, most suitable offset is optionally selected as an inconsistent signal according to preparatory condition, to compensate the error of reconstructed image data after coding.
Description
Technical field
The present invention relates to a kind of image processing apparatus and method;More particularly to a kind of selective pixel value difference compensation method
With device.
Background technology
In the video coding in modern age(H.264, HEVC, AVS), contain wave filter in a loop(in-loop
filter), the picture encoded is got rid of into partial noise, such as get rid of blocking artifact(blocking effect)'s
Then it is referred to as deblocking filter(De-blocking filter).And the result being filtered, follow-up input can be stored for
Video pictures in coding when it is referenced, so the wave filter referred to as in loop.
But wave filter is often predicted only with reference to the view data after coding in loop, such as inter-picture prediction
(Inter-prediction).Because the numerical value that the foundation of filter reference in only loop is only prediction does not refer to original image
Data, therefore can not accurately correct the difference encoded between picture and raw frames.
The content of the invention
An object of the present invention is providing a kind of selective pixel value difference compensation method and device, and this method is with reference to volume
View data before code and the view data after coding, are produced multiple according to pixel difference state in the square of two view data
Offset, optionally selects most suitable offset as an inconsistent signal, to compensate weight after coding according to preparatory condition
Build the error of view data.
One embodiment of the invention provides a kind of selective pixel value difference compensation method, includes the following steps:It is first
First, receive reconstructed image data after reconstructed image data after a coding, coding and divide into multiple squares.Then, one is received to encode
Preceding view data.Afterwards, according to each square pixel before encoding view data with coding after reconstructed image data identical bits
The difference status put, statistical discrepancy state produces multiple offsets, optionally chooses an offset and is differed with to produce one and is write
Number.
Another embodiment of the present invention provides processing unit in a kind of loop, it is adaptable to a coded system, includes one
Difference compensation device.Difference compensation device receives reconstructed image data and the preceding view data of a coding after a coding, and after coding
The picture of reconstructed image data divides into multiple squares.Pixel before encoding picture number of the difference compensation device according to each square
According to the difference status with reconstructed image data same position after coding, an inconsistent signal is produced.Wherein, difference compensation device
The difference status of each square comprising the edge of reconstructed image data after coding, with multiple classification numerical value of angle change state,
Multiple brightness variable condition section numerical value, optionally select the part classifying numerical value of multiple classification numerical value and the plurality of brightness
The partial sector numerical value of variable condition section numerical value produces the inconsistent signal as offset.
Another embodiment of the present invention provides a kind of coded system, includes a motion compensating device and a difference
Compensation device.Motion compensating device is to compensate reconstructed image data and input picture at present after previous coding(Current volume
View data before code)Between displacement error.Difference compensation device is coupled between conversion quantization device and motion compensating device, is connect
Receive reconstructed image data and view data before current encoder after current encoder.Wherein, after current encoder reconstructed image data picture
Multiple squares are divided into face, difference compensation device according to each square pixel before current encoder view data and current encoder
Multiple difference status of reconstructed image data same position, produce an inconsistent signal afterwards.Wherein, difference compensation device selectivity
Ground selects the numerical value of the part variation state of multiple difference status as offset, produces above-mentioned inconsistent signal.
The pixel value difference compensation device of the selectivity of the embodiment of the present invention and processing unit in method, loop and coding system
Reconstructed image data and view data before coding after system, reference encoder, and a variety of moulds are used to each square in units of square
Formula is assessed, and judges which kind of pattern picture quality can be improved under conditions of not increasing or increasing a little bit rate using.And compare
The ratio of the incrementss of special rate is smaller than the ratio of the picture quality improved in one embodiment.
Brief description of the drawings
Figure 1A is the schematic diagram for showing coded system according to an embodiment of the invention.
Figure 1B is the schematic diagram for showing coded system according to another embodiment of the present invention.
Fig. 1 C are the schematic diagram for showing selective pixel value difference compensation device according to an embodiment of the invention.
Fig. 2A is the flow chart for showing selective pixel value difference compensation method according to an embodiment of the invention.
Fig. 2 B are the flow chart for showing step S203 according to an embodiment of the invention.
Fig. 3 A are display spike according to an embodiment of the invention and the schematic diagram of low ebb.
The schematic diagram that Fig. 3 B classify for display angle according to an embodiment of the invention.
The schematic diagram that Fig. 3 C are distributed for display section according to an embodiment of the invention.
【Symbol description】
100 coded systems
101 difference compensation devices
102 coding controllers
103 conversion quantization devices
103a conversions/scaling/quantifying unit
103b is scaled and investigation converting unit
Unit is preset in 104 pictures
105 motion compensations and estimation unit
105a motion compensation units
105b moves estimation unit
106 unrest degree encoders
107 remove block effect wave filter
101a lines mode processing units
101b smooth mode processing units
101c judging units
Embodiment
Figure 1A shows the schematic diagram of the coded system 100 of one embodiment of the invention.Coded system 100 includes a selectivity
Pixel value difference compensation device(Selective Pixel DisparityCompensation, SPDC)101st, a coding-control
Device(Coder control)102nd, a conversion quantization device 103, an intra-frame prediction unit(Intra-frame
Prediction)104th, a motion compensation and the unrest degree encoder of estimation unit 105 and one(Entropy Coding)106.One
Embodiment, conversion quantization device 103 includes one conversion/scaling/quantifying unit(Transform/Scal./Quant.)103a、
And scaling and investigation converting unit(Scaling&Inv.Transform)103b.One motion compensation includes with estimation unit 105
One motion compensation unit(Motion Compensation)The movement estimation units of 105a and one(Motion Estimation)
105b。
Coded system 100 is received before coding(Input)View data BCD, will be produced before coding after view data BCD codings
Reconstructed image data ACD after coding, then after the processing such as motion compensation, via the random output image data OS of degree encoder 106,
To export to code translator decoding and display image.In one embodiment, decoding apparatus for example can be to be arranged at a TV
(TV)Or the device of various displayable images(Such as contact surface plate, hand-held moving device, mobile glasses)Decoding apparatus.
It may be noted that in one embodiment, in the figure, signal or data symbol CD represent a control data(Control
Data), QTC represent a quantization transmission coefficient(Quant.Transf.coeffs), OV represent output image(Output Video)
Mobile data is represented for coding side reference, MD(Motion Data).
In one embodiment, the coupling of difference compensation device 101 conversion quantization device 103 and motion compensation unit 105a.
Difference compensation device 101 receives reconstructed image data ACD and view data BCD before coding after coding, and view data is divided into
Multiple squares(Blocks), for each square before encoding, the state difference of the pixel of rear view data BCD, ACD, calculate
Go out offset Com(It is not shown), to produce inconsistent signal(Disparity Signal)DS to random degree encoder 106 there is provided
Extra information improves the quality of image.
It may be noted that the difference compensation device 101 of the embodiment of the present invention is different from known technology, difference compensation device 101 is another
It is outer according to coding before raw image data compensate, it is non-as known technology only with reference to prediction data as offset according to
According to.Furthermore, difference compensation device 101 can optionally be found out most suitable according to the pixel classifications of each square with state characteristic
Offset Com, is used as inconsistent signal DS.
It may be noted that the difference compensation device 101 of the present invention may be disposed in various coded systems, however it is not limited to above-mentioned implementation
The framework of example.Furthermore, in addition to the difference compensation device 101 of the present invention, those skilled in the art, it should be appreciated that above-described embodiment
Framework operation principles and method, repeat no more its content.
In one embodiment, as shown in Figure 1B, coded system 100 also removes block effect wave filter comprising one(De-
blocking Filter)107.The difference compensation device 101 of the present embodiment may be disposed at after block effect wave filter 107,
The output signal and view data BCD before coding for removing block effect wave filter 107 are received, to export inconsistent signal DS to unrest degree
Encoder 106.
In one embodiment, when coded system 100 is not provided with block effect wave filter 107, difference compensation device 101
Processing unit in loop can be used as(In-loop processing device)Or wave filter in loop(In-loop filter)
Use, block effect wave filter 107 is removed in substitution.
Fig. 1 C show the schematic diagram of processing unit in the loop of one embodiment of the invention-difference compensation device 101.Difference
Compensation device 101 receives reconstructed image data ACD and view data BCD before coding after coding, according to each party in view data
The pixel of block before encoding view data BCD with coding after reconstructed image data ACD same position pixels diversity factor, produce one
Inconsistent signal DS.
As shown in Figure 1 C, difference compensation device 101 includes a lines pattern(Texture Mode)Processing unit 101a, one
Smooth mode(Flat Mode)A processing unit 101b and judging unit 101c.
Lines mode processing units 101a is by each pixel basis edge of each squares of reconstructed image data BCD after coding
(Edge)With height(Peak and Valley)State classification, the number of pixels for calculating each classification produces a differential counting
Value count_t.After calculation code in reconstructed image data ACD each classification pixel value summation, produce and classify total after a coding
With value sum(Coded)_t.Before calculation code in view data BCD each classification pixel value summation, produce classification before coding total
With value sum(Org)_t;The pixel of each classification is calculated using random degree encoder 106, the bit rate of the classification is obtained(Bit
rate).By the total value sum that classifies after the differential counting value count_t of each classification, the coding of each classification(Coded)_t、
Classify total value sum before the coding of each classification(Org)_ t and the bit rate of the classification substitute into a bit rate and distortion value side
Formula, and default offset is substituted into, to calculate the first bit rate and distortion value RDt and find out the offset of each classification
(Compensation).Then, selection produces minimal bit rate and distortion value(Rate-Distortion)The one of RDt is optimal
Offset is used as the first offset Com1.
Smooth mode processing unit 101b is bright by whole pixel basis of reconstructed image data BCD each square after coding
Multiple brightness sections are divided into the distribution of degree, calculate the number of pixels of each section, produce a sector counter values count_f;Meter
The pixel value summation of each section in reconstructed image data ACD after coding is calculated, section total value sum after coding is produced(Coded)_
f;Before calculation code in view data BCD each section pixel value summation, produce coding front section total value sum(Org)_f;
The bit rate of each section pixel is calculated using random degree encoder 106;Then, in multiple sections, the continuous multiple sections of selection,
By section total value sum after the sector counter values count_f of each section, coding in the plurality of section(Coded)Before _ f, coding
Section total value sum(Org)_ f and the bit rate substitute into a bit rate and distortion value equation, and substitute into default compensation
Value, to calculate the second bit rate and distortion value RDf and find out the offset of each section(Compensation).Then, select
The optimal offset for producing minimal bit rate and distortion value RDf is selected as the second offset Com2.
Judging unit 101c compares the first bit rate and distortion value RDt and the second bit rate and distortion value RDf, when the first ratio
When special rate is less than zero less than the second bit rate and distortion value RDf, and the first bit rate with distortion value RDt with distortion value RDt, output
First offset Com1, to be used as inconsistent signal DS.And when the second bit rate is less than the first bit rate with distortion value RDf and loses
True value RDt, and the second bit rate and distortion value RDf are when being less than zero, the second offset Com2 are exported, to be used as inconsistent signal
DS.It may be noted that bit rate is the smaller the better with distortion value, and is necessarily less than zero, just image quality can be lifted in the case where not influenceing efficiency.
When first, second bit rate and distortion value RDt, RDf are all higher than being equal to zero, using close pattern(Off Mode), not to compiling
Reconstructed image data ACD is compensated after code, and reconstructed image data ACD is exported after directly encoding.
As shown in Figure 1A, 1B, the random degree encoder 106 of coded system 200 receives inconsistent signal DS, is write according to differing
Reconstructed image data ACD after number compensation coding, to export after the coding after compensation reconstructed image data to a decoder.
Next, describe in detail the present invention selectivity pixel value difference compensation method, please also refer to Fig. 1 C, 2A, 2B,
3A, 3B and 3C.
As shown in Fig. 2A, 2B, the flow chart of the pixel value difference compensation method of Fig. 2A, 2B display present invention selectivity.The party
Method is comprised the steps of:
Step S201:Start.
Step S202:By the reconstruction image picture after coding(Reconstructed image data ACD after encoding)Cut into N × N(Example
Such as 6464) square, difference compensation device 101 receive N × N square.Wherein, N is positive integer.
Step S203:This square and uncoded preceding image frame same position are found for each square, identical bits are calculated
The diversity factor of the reconstruction image picture of the pixel put in encoded and uncoded preceding image frame, optionally finds out various spies
Property classification under values of disparity between pixel, and select most suitable values of disparity, inconsistent signal produced according to values of disparity
DS, to compensate the diversity factor(disparity compensation).Wherein, the generation of diversity factor can be to each square
Using a variety of(Such as Three models)Go to assess, this Three models includes a lines pattern(Texture Mode), it is one smooth
Pattern(Flat Mode), a close pattern(OFF Mode), to determine which kind of pattern reasonable bit rate can be being increased using
Under the conditions of increase picture quality.In addition, this increased bit of inconsistent signal institute for additionally sending out, is set as not resulting in same
The size for making bit rate increase under conditions of sample picture quality, for example(Bjontegaard Delta Bit-Rate)BDBR<0.
Fig. 2 B are refer to, step S203 is also comprised the steps of:
Step S2031:One lines pattern, lines schema processing unit 101a, with edge are provided(edge)Concept
By pixel(pixel)Classification, such as be following five kinds of states by pixel classifications:(1)Edge and for low ebb(valley)Pixel,
(2)Non-edge and for low ebb pixel,(3)Edge and for spike(peak)Pixel,(4)Non-edge and for spike(peak)Pixel,
(5)It is all non-above.Then, it is right(1)~(4)The pixel of kenel is counted, and finds out the quantity of each kenel group(count), after coding
Pixel value sum, coding preceding pixel value sum, to find out the offset that this four kenel group is best suitable in single square.It may be noted that
The spike at edge(Peak)And low ebb(Valley), as shown in Figure 3A.
In one embodiment, as shown in Figure 3 B, due to conventional angle have level, vertical, Oblique 45 Degree, oblique 135 degree,
So can be the style of this four kinds different angles in lines mode step S2031(pattern), it is out of the ordinary to count, it is each by what is found out
The quantity of kenel group(Count), pixel value sum, coding preceding pixel value sum substitute into lines mode processing units 101a after coding
A bit rate and distortion value(RD)Equation:
According to the size of bit rate and distortion value RDt, which angle is judged for best angle, and can obtain the benefit that best angle is calculated
Repay value Com(Compensations).
First, such as Fig. 3 B(A)Style 1 shown in, when pixel detection angle be horizontal spike or low ebb when, now arrange in pairs or groups
The edge of vertical direction, pixel is divided into(1)Vertical edge and horizontal spike,(2)Non-perpendicular edge and horizontal spike,(3)Hang down
Straight edge and horizontal low ebb,(4)Non-perpendicular edge and horizontal low ebb.Then, each pixel in the square is searched, to find out this
The quantity of four kinds of classification difference(Count), calculating this four kinds classification, reconstructed image data ACD square is whole in encoded respectively
Pixel value sum(sum(Coded)), calculate this four kinds classification whole pixel values of view data BCD square before encoding respectively
Sum(sum(Org)).Then, by each classification i(I.e.(1)、(2)、(3)、(4))Count, sum(Coded)、sum
(Org)Equation RDt is substituted into, default offset Com is utilized(compesations)Bit rate and distortion value RDt are calculated,
To find out most suitable offset.For example, in style 1,(1)The Count values that vertical edge is calculated with horizontal spike
For 100 pixels, it is 3 that it, which substitutes into and obtains the offset to be sent after the predesigned compensation value test of equation,;(2)Non-perpendicular edge with
The Count values of horizontal spike are 5, and it is 0 that it, which substitutes into and obtains the offset to be sent after the predesigned compensation value test of equation,;(3)Hang down
Straight edge is 100 with the Count values that horizontal low ebb is calculated, and it obtains what is sent after substituting into the predesigned compensation value test of equation
Offset is 2;(4)Non-perpendicular edge is 10 with the Count values that horizontal low ebb is calculated, and it substitutes into the predesigned compensation value of equation
It is 1 that the offset to be sent is obtained after test.Then, the RDt values obtained by being substituted into due to offset 3 after equation are minimum, therefore
Select offset 3.
It may be noted that in one embodiment, in the classification of style 1(1)Vertical edge and horizontal spike with(3)Vertical edge with
In the state of horizontal low ebb, in order to allow vertical edge to obtain preferably offset, it can allow to mingle from positive and negative number
(Sign)Or be entirely positive sign(Unsign)Offset;Relatively, in the classification of style 1(2)Non-perpendicular edge and horizontal spike
With(4)In the state of non-perpendicular edge and horizontal low ebb, in order to obtain relatively low bit rate, then not allowing to select has sign
Offset, can only select positive number(Unsign)Offset.In another embodiment, those skilled in the art, can the rest may be inferred
Determine that the classification setting compensation value of style 2~4 selects mode.
Such as Fig. 3 B(B)Style 2 shown in, when pixel detection direction be vertical spike or low ebb when, level of now arranging in pairs or groups
To edge, pixel is divided into(1)Horizontal edge and vertical spike,(2)Non-horizontal edge and vertical spike,(3)Horizontal edge with
Vertical low ebb,(4)Non-horizontal edge and vertical low ebb.Then, each pixel in the square is searched, to find out this four kinds classification
Quantity(count), whole pixel value sums of each classification of the square in reconstructed image data ACD after calculation code(sum
(Coded)), whole pixel value sums of each classification of the square in view data BCD before calculation code(sum(Org)).Connect
, an offset is calculated with above-mentioned style 1.
Such as Fig. 3 B(C)Style 3 shown in, when pixel detection direction be 45 degree of spikes or low ebb when, level of now arranging in pairs or groups
Or vertical direction edge, pixel is divided into(1)Horizontally or vertically edge and 45 degree of spikes,(2)Non-horizontal or vertical edge and 45
Degree spike,(3)Horizontally or vertically edge and 45 degree of low ebbs,(4)Non-horizontal or vertical edge and 45 degree of low ebbs.Then, searching should
Each pixel in square, to find out the quantity of this four kinds classification(count), should in reconstructed image data ACD after calculation code
The whole pixel value sums of each classification of square(sum(Coded)), before calculation code in view data BCD the square it is each
The whole pixel value sums of classification(sum(Org)).Then, an offset is calculated with above-mentioned style 1.
Such as Fig. 3 B(D)Style 4 shown in, when pixel detection direction be 135 degree of spikes or low ebb when, level of now arranging in pairs or groups
Or vertical direction edge, pixel is divided into(1)Horizontally or vertically edge and 135 degree of spikes,(2)Non-horizontal or vertical edge with
135 degree of spikes,(3)Horizontally or vertically edge and 135 degree of low ebbs,(4)Non-horizontal or vertical edge and 135 degree of low ebbs.Then, search
The each pixel sought in the square, to find out the quantity of this four kinds classification(count), reconstructed image data ACD after calculation code
In the square the whole pixel value sums of each classification(sum(Coded)), the square in view data BCD before calculation code
Each whole pixel value sums of classification(sum(Org)).Then, an offset is calculated with above-mentioned style 1.
In this way, lines mode processing units 101a can find out four kinds of angles in the direction of element detection(Level, it is vertical,
45 degree, 135 degree)Under the best angle for obtaining most suitable offset, optionally send out best angle style numbering and its
The four offset Com possessed(Compensations).For example, it is assumed that best angle is style(2)And four offset difference
For 0,1,3,2.
Step S2032:One smooth mode is provided(Flat Mode), smooth mode processing unit 101b is with length in this step
Bar statistical chart(histogram)Concept, the pixel of each square is divided into multiple brightness sections, is that each section finds out group's
Quantity(count), it is every in view data BCD before the pixel value sum of each section, coding in reconstructed image data ACD after coding
The pixel value sum of one section.For there is the section of statistic in single square, the compensation being best suitable in single square is found out
Value.
In one embodiment, as shown in Figure 3 C, brightness is divided into 32 ranks by smooth mode processing unit 101b(The highest of pixel
Brightness 256 divided by 8, divisor 8 is any setting numerical value, and the invention is not restricted to the square of this size and this divisor).Then, to one
Square is counted, and is progressively counted to Bin32 by the 0 of the binary numeral Bin of brightness.It note that binary system in the figure
The side of numerical value Bin3,6,7,8,12,14,15 has cumulative statistics numerical value, is that these Bin for having cumulative statistics numerical value redefine ground
Location, is newly defined as Bin0,3,4,5,9,11,12 respectively.Smooth mode processing unit 101b is in having the bin of value, the company of selecting
It is continuous 4,0,3,4,5 or 3,4,5,9 or 4,5,9,11 in such as legend, one group is found out in following bit rate and distortion value equation
RDf...(2)In, the combination of optimum value can be obtained.Assuming that the RDf values of original Bin6,7,8,12 be 500, than original Bin3,
6th, 7,8 RDf values 1000 are low, then smooth mode processing unit 101b can select have value by continuous 4 Bin6
Bin6,7,8,12, Bin6 order is re-addressed as Bin3.Then, by the starting Bin of the combination of optimum value binary number
Value index(Bin index)And four Com2(Compensations)Coding, in one embodiment for example(3;- 1, -2,1,2),
Send out, wherein 3 be index, -1, -2,1,2 is four Com2 codings.It may be noted that sending out three ropes that index Bin3 represents to continue
It is sequentially Bin4, Bin5, Bin9 to draw, and Bin4, Bin5, Bin9 also have each self-corresponding four Com2 values, this four Com2 values
- 1, -2,1,2 are may be selected from one embodiment.
Step S2033:Judge whether RDt is less than RDf, if so, into step S2034;If it is not, into step S2035.
Step S2034:Judge whether RDt is less than 0, if so, into step S2036;If it is not, into step S2037.Its
In, when RDt value is less than 0, represent that adding this offset can benefit to picture quality lifting;And when RDt value is more than 0
When, represent that add this offset is no advantage to picture quality lifting.
Step S2035:Judge whether RDf is less than 0, if so, into step S2038;If it is not, into step S2037.Its
In, when RDf value is less than 0, represent that adding this offset can benefit to picture quality lifting;And when RDf value is more than 0
When, represent that add this offset is no advantage to picture quality lifting.
Step S2036:The offset Com1 gone out using lines mode computation produces inconsistent signal DS.
Step S2037:Using close pattern, any offset of not bothering to see me out maintains the state after this block encription.
Step S2038:The offset Com2 gone out using smooth lines mode computation produces inconsistent signal DS.
Step S204:Terminate.
It note that known technology video coding HEVC sampling is adaptively offset(Sample Adaptive Offset,
SAO)In, it is divided into close pattern(Off Mode), edge difference(Edge Offset, EO), frequency range difference(Band Offset,
BO)Three models, wherein, EO patterns include low ebb(Valley), it is recessed(concave), it is convex(convex), spike(Peak)Four kinds
Pattern, in this statistical smoothly will be integrally incorporated in marginal zone, not to edge(Edge)Especially processing, furthermore also without
Limit and the number of having is selected under which state(Sign), which is from nothing number(Unsign)Offset;And BO statistical, then not
Processed specifically designed for the section for having statistic, no matter the section for whetheing there is statistic can be used.And the embodiment of the present invention, then
Most suitable statistics section can be optionally found out, to produce most suitable offset.
In coding, the inconsistent signal DS that the pixel value difference compensation device 101 of selectivity of the invention is produced includes
Mode index(Mode index), style index(Pattern index)And offset Com(Compensations).
In one embodiment, mode index can be:Close pattern(0), lines pattern(1), smooth mode(2).Due in video pictures,
Relevance presence is had between adjacent picture, smooth area or lines area can be continuously a piece of, so in coding, content can be used
The random degree coding of adaptability(context adaptive entropy coding)Mode, utilizes the mode index of previous square
To predict the mode index of current square, increase coding efficiency.And two of the style index and smooth mode in lines pattern enter
Numerical index processed(Bin index), each pattern offset, can be used lossless coding(lossless)Mode encode.Need note
Meaning, it will be understood by a person skilled in the art that above-mentioned lossless signal coding and lossless coding mode, are repeated no more.
After the pixel value difference compensation device and method of the selectivity of the embodiment of the present invention, reference encoder reconstructed image data with
View data before coding(Original image)And in units of square, each square is assessed using various modes, what judgement uses
The pattern of kind can increase picture quality under conditions of bit rate is not increased.
Claims (19)
1. a kind of processing unit in loop a, it is adaptable to coded system, wherein, processing unit is included in the loop:
One difference compensation device, the difference compensation device receives reconstructed image data and picture number before a coding after a coding
According to the picture of reconstructed image data divides into multiple squares after the coding, and the difference compensation device is according to each side
The difference status of the pixel of block view data and the same position of reconstructed image data after the coding before the coding, is produced
One inconsistent signal, with according to reconstructed image data after being encoded described in the inconsistent signal compensation;
Wherein, the difference status of each square of the difference compensation device includes reconstruction image number after the coding
According to edge and angle change state multiple classification numerical value, multiple brightness variable condition section numerical value, the difference compensation dress
Put the part classifying numerical value and the multiple brightness variable condition section numerical value for optionally selecting the multiple classification numerical value
Partial sector numerical value produces the inconsistent signal as offset.
2. processing unit in loop according to claim 1, wherein, the difference compensation device is coupled to the coding system
Between one motion compensation unit of system and a quantization converting unit.
3. processing unit in loop according to claim 1, wherein, a unrest degree encoder of the coded system is according to institute
Reconstructed image data after being encoded described in inconsistent signal compensation is stated, reconstructed image data after the coding after compensation is exported
To a decoder for decoding.
4. processing unit in loop according to claim 1, wherein, the difference compensation device is included:
One lines mode processing units, the lines mode processing units are by each pixel basis edge of the square and just
Dotted state is classified, and is calculated the number of pixels of each classification and is produced a differential counting value;Calculate reconstructed image data after the coding
In each classification pixel value summation, produce total value of classifying after a coding;Calculate every in view data before the coding
The pixel value summation of the individual classification, produces classification total value before coding;Each classified pixels are encoded by a unrest degree
Device calculates the bit rate of the classification;Classification total value after the differential counting value, the coding according to each classification,
Total value of classifying before the coding and the bit rate produce the offset and a bit rate and the distortion of each classification
Value, according to offset described in the size selection one of the bit rate and distortion value as one first offset, and described first mends
It is one first bit rate and distortion value that the corresponding bit rate of value, which is repaid, with distortion value.
5. processing unit in loop according to claim 4, wherein, the difference compensation device is included:
Whole pixels of the square are divided into multiple brightness sections by one smooth mode processing unit, calculate each area
The number of pixels of section, produces a sector counter values;Calculate the pixel of each section in reconstructed image data after the coding
It is worth summation, produces section total value after a coding;The pixel value for calculating each section in view data before the coding is total
With the coding front section total value of generation one;The area is calculated by the random degree encoder using the pixel of each section
The bit rate of section;The continuous multiple sections of selection, according to sector counter values of each section in the multiple section,
Section total value, the coding front section total value and the bit rate produce the described of each section after the coding
Offset and a bit rate and distortion value, one the is used as according to offset described in the size selection one of the bit rate and distortion value
Two offsets, and the bit rate and distortion value of second offset are the second bit rate and distortion value.
6. processing unit in loop according to claim 5, wherein, the difference compensation device is included:
One judging unit, relatively more described first bit rate and distortion value and second bit rate and distortion value, when described first
When bit rate is less than zero less than second bit rate and distortion value, and first bit rate with distortion value with distortion value, institute
State judging unit and regard first offset as an inconsistent signal output;When second bit rate and distortion value are less than institute
State the first bit rate and distortion value, and second bit rate with distortion value when being less than zero, the judging unit is by described second
Offset is used as the inconsistent signal output.
7. processing unit in loop according to claim 5, wherein, the inconsistent signal includes mode index, style
Index or binary numeral index and first offset or second offset.
8. processing unit in loop according to claim 1, wherein, the difference compensation device removes block effect located at one
The rear end of wave filter, for receiving the output signal for removing block effect wave filter.
9. a kind of selective pixel value difference compensation method, comprising:
Receive reconstructed image data after reconstructed image data after a coding, the coding and include multiple squares;
Receive view data before a coding;And
Pixel according to each square before the coding view data with after the coding reconstructed image data it is identical
The difference status of position, counts the difference status to produce multiple offsets, optionally choose an offset with
An inconsistent signal is produced, with according to reconstructed image data after being encoded described in the inconsistent signal compensation.
10. method according to claim 9, wherein, the difference status includes lines difference and smoothness difference.
11. method according to claim 10, wherein, the lines difference is the point using multiple angles and edge formation
Peak and low ebb and classified, count multiple offsets of each classification, calculate each described using a default equation
The bit rate and distortion value of offset, select one first offset according to the multiple bit rate and distortion value produce it is described not
Consistent signal, and the corresponding bit rate of first offset and distortion value are one first bit rate and distortion value.
12. method according to claim 11, wherein, the smoothness difference divides into the Luminance Distribution of the square
Multiple sections, count multiple offsets of each section, and each offset is calculated using a default equation
Bit rate and distortion value, select one second offset according to the multiple bit rate and distortion value and are write to produce described differ
Number, and the corresponding bit rate of second offset and distortion value are one second bit rate and distortion value.
13. method according to claim 12, wherein, methods described also include will first bit rate and distortion value and
Second bit rate is compared with distortion value, is determined according to comparative result by first offset and second compensation
Which in value is used as the inconsistent signal output.
14. method according to claim 13, wherein, first bit rate and distortion value and second bit rate with
When distortion value is less than zero, the inconsistent signal is exported.
15. method according to claim 13, wherein, when first bit rate and distortion value and second bit rate
When being more than or equal to zero with distortion value, the inconsistent signal is not exported.
16. a kind of coded system, comprising:
One motion compensating device, the mobile mistake to compensate reconstructed image data and view data before current encoder after previous coding
Difference;And a difference compensation device, the motion compensating device is coupled, reconstructed image data and institute after the current encoder is received
View data before current encoder is stated, wherein, the picture of reconstructed image data divides into multiple squares after the current encoder, described
The pixel of each square of difference compensation device foundation view data and weight after the current encoder before the current encoder
Multiple difference status of the same position of view data are built, an inconsistent signal are produced, with according to the inconsistent signal compensation
Reconstructed image data after the current encoder;
Wherein, the difference compensation device optionally selects the numerical value conduct of the part variation state of the multiple difference status
Offset, produces the inconsistent signal.
17. system according to claim 16, wherein, the system also includes a de-blocking filtering device, to filter out
The block effect of reconstruction image after current encoder is stated, and the difference compensation device receives the output of the de-blocking filtering device.
18. system according to claim 16, wherein, the difference compensation device is coupled to a conversion quantization device and institute
State between motion compensating device.
19. system according to claim 16, wherein, the system also includes a unrest degree encoder, the random degree coding
Device is according to reconstructed image data after current encoder described in the inconsistent signal compensation.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310236938.9A CN104243996B (en) | 2013-06-14 | 2013-06-14 | The pixel value difference compensation method of processing unit and selectivity and coded system in loop |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201310236938.9A CN104243996B (en) | 2013-06-14 | 2013-06-14 | The pixel value difference compensation method of processing unit and selectivity and coded system in loop |
Publications (2)
Publication Number | Publication Date |
---|---|
CN104243996A CN104243996A (en) | 2014-12-24 |
CN104243996B true CN104243996B (en) | 2017-07-28 |
Family
ID=52231213
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201310236938.9A Active CN104243996B (en) | 2013-06-14 | 2013-06-14 | The pixel value difference compensation method of processing unit and selectivity and coded system in loop |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN104243996B (en) |
Families Citing this family (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110910805A (en) * | 2019-12-31 | 2020-03-24 | 深圳创维-Rgb电子有限公司 | Compensation adjusting device and compensation adjusting method of display screen |
CN111654704A (en) * | 2020-05-26 | 2020-09-11 | 新相微电子(上海)有限公司 | Automatic multi-mode image compression control method and device |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1154049A (en) * | 1995-12-29 | 1997-07-09 | 大宇电子株式会社 | Method for rectifying channel errors in image signal transmitted from block-based encoder |
CN1728182A (en) * | 2004-06-29 | 2006-02-01 | 英特尔公司 | Image edge filtering |
Family Cites Families (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20080298472A1 (en) * | 2007-06-04 | 2008-12-04 | Texas Instruments Incorporated | Throughput Performance When Applying Deblocking Filters On Reconstructed Image Frames |
-
2013
- 2013-06-14 CN CN201310236938.9A patent/CN104243996B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN1154049A (en) * | 1995-12-29 | 1997-07-09 | 大宇电子株式会社 | Method for rectifying channel errors in image signal transmitted from block-based encoder |
CN1728182A (en) * | 2004-06-29 | 2006-02-01 | 英特尔公司 | Image edge filtering |
Also Published As
Publication number | Publication date |
---|---|
CN104243996A (en) | 2014-12-24 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
US9204165B2 (en) | Method and apparatus for decoding intra prediction mode | |
RU2710996C2 (en) | Method of reconstructed unit formation | |
CN103081474B (en) | For the device that motion picture is decoded | |
TWI661715B (en) | Method of decoding video data | |
US20150110185A1 (en) | Method and apparatus for encoding intra prediction information | |
CN102420988B (en) | Multi-view video coding system utilizing visual characteristics | |
US9813708B2 (en) | Method and system of coding prediction for screen video | |
CN107105252A (en) | The method and apparatus for deriving intra prediction mode | |
CN108366260A (en) | Equipment for generating prediction block in intra prediction | |
CN106937116B (en) | Low-complexity video coding method based on random training set adaptive learning | |
CN104662908A (en) | Depth map coding | |
CN105338352A (en) | Adaptive dequantization method and device in video decoding | |
CN104883565A (en) | Decision-making method and device for intra-frame prediction mode of high efficiency video coding | |
CN108141593A (en) | For be directed to the efficient intraframe coding of deep video based on the discontinuous method of depth | |
CN103596004A (en) | Intra-frame prediction method and device based on mathematical statistics and classification training in HEVC | |
CN101854548A (en) | Wireless multimedia sensor network-oriented video compression method | |
CN106303521B (en) | A kind of HEVC Rate-distortion optimization method based on sensitivity of awareness | |
CN105898300A (en) | Improved transformation coefficient sign data hiding method based on recovered transformation coefficient | |
JP2021090135A (en) | Signal processing device and signal processing method, system, learning method, and program | |
CN103442229B (en) | The bit rate estimation method of SAO mode adjudgings suitable for the encoder of HEVC standard | |
CN104243996B (en) | The pixel value difference compensation method of processing unit and selectivity and coded system in loop | |
CN102065293B (en) | Image compression method based on space domain predictive coding | |
CN100534182C (en) | Method for coding a video image taking into account the part relating to a component of a movement vector | |
CN1317125A (en) | Method and device for estimating motion in digitized image with pixels | |
CN108174201A (en) | Terminate the video-frequency compression method of encoding block coding mode search in advance by video content |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
GR01 | Patent grant | ||
GR01 | Patent grant |