CN101848394A - Method for optimizing AVS coding power consumption model of radio video sensor - Google Patents
Method for optimizing AVS coding power consumption model of radio video sensor Download PDFInfo
- Publication number
- CN101848394A CN101848394A CN 201010184247 CN201010184247A CN101848394A CN 101848394 A CN101848394 A CN 101848394A CN 201010184247 CN201010184247 CN 201010184247 CN 201010184247 A CN201010184247 A CN 201010184247A CN 101848394 A CN101848394 A CN 101848394A
- Authority
- CN
- China
- Prior art keywords
- macro
- current
- frame image
- macro block
- coding
- 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
Images
Landscapes
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The invention discloses a method for optimizing an audio video standard (AVS) coding power consumption model of a radio video sensor. The method comprises the following steps of: directly performing AVS coding on the image of a frame I; dividing prediction modes for the image of the frame P into a skip mode, an inter-frame 16*16 prediction mode and rest inter-frame prediction modes; determining macro blocks using the skip mode and the inter-frame 16*16 prediction mode according to listed power consumption models when a target bit rate and available coding power consumption are known; and selecting the best prediction mode for the rest macro blocks from the rest inter-frame prediction modes apart from the skip mode and the inter-frame 16*16 prediction mode by using Lagrange cost function so as to ensure low computational complexity and the validity of the power consumption model. The method can adjust the number of the macro blocks using the skip mode, the inter-frame 16*16 prediction mode and the rest inter-frame modes in the image of the frame in a video signal according to the listed power consumption models in real time when the available coding power consumption reduces continuously, so that the quality of the video image of the radio video sensor keeps the best under the condition that the energy is limited.
Description
Technical field
The present invention relates to a kind of optimisation technique of video coding power consumption, especially relate to a kind of AVS (Audio Video Coding Standard, Audio Video coding Standard) coding power consumption model optimization method of radio video sensor.
Background technology
AVS (Audio Video coding Standard, digital audio/video encoding and decoding technique standard) be that Chinese first has the digital audio/video encoding and decoding standard of independent intellectual property right, its code efficiency and competitive H.264/MPEG-4AVC suitable with current international standard, but be better than H.264/MPEG-4AVC at implementation complexity and high definition application facet.
Radio video sensor generally adopts powered battery, no longer changes battery after having disposed usually, if encoder is according to traditional method for video coding, then after the energy content of battery is reduced to certain limit, the video image quality of coding will be poor make us and can't accept.For the service efficiency that makes radio video sensor reaches the highest, study its power consumption model, make under the situation of its energy limited, video image quality reaches optimum and just seems and be necessary very much.Video coding and transfer of data are two main aspects of the power consumption of radio video sensor, studies show that video sequence (as QCIF176 * 144) for small in resolution, the video coding power consumption can account for 2/3 of radio video sensor total power consumption, for high-resolution video sequence, the video coding power consumption of radio video sensor is higher.And from the power consumption of radio video sensor, the influence of video coding has dual character: on the one hand, video coding compression efficiently can reduce the bit number of transfer of data significantly, can satisfy the demand of real-time video communication preferably; On the other hand, video coding compression efficiently often needs the calculating power consumption of higher computation complexity and Geng Gao, and this is a challenge to power limited wireless video sensor.
At present, the research for the video coding power consumption be primarily aimed at H.26x, international standards such as MPEG (Moving Pictures ExpertsGroup, dynamic image expert group)-2, MPEG-4.These researchs are mainly by analyzing each module of coding, thereby set up total coding side power consumption model.As the scale-model investigation of the disclosed wireless mobile environment video coding dynamic power consumption of sensing technology journal (2009 (3) 354-357), drawn the overall power consumption model of coding side from frame mode coding power consumption and coded in inter mode power consumption analysis, frame refreshing rate parameter beta=1/T (an I frame just being arranged) and quantization parameter have been introduced every the T-1 frame, the overall power consumption model has been obtained comparatively desirable video coding power consumption control effect by this two parameter controls and adjusting.But this method is not considered the relation between power consumption, code check and the distortion, though when the frame refreshing rate is high, encoder complexity is lower, but this moment, the code check of coding also can be very high, band-limited wireless channel is difficult to so high code check of transmission, and the frame refreshing rate is when very high, such as an I frame is just arranged every 2 frames, can produce very big time redundancy during coding, code efficiency is very low.
Summary of the invention
Technical problem to be solved by this invention provides a kind of radio video sensor that can make under the situation of energy constraint system, and it is minimum that the video distortion of coding reaches, and the low AVS coding power consumption model optimization method of computation complexity.
The present invention solves the problems of the technologies described above the technical scheme that is adopted: a kind of AVS coding power consumption model optimization method of radio video sensor may further comprise the steps:
1. the initial target bit rate of the current vision signal of obtaining of setting wireless video sensor, be designated as R, wherein, vision signal comprises a plurality of image sets, each image sets mainly is made up of an I two field picture and at least one P two field picture, the start frame image of I two field picture during for each image sets coding, the P two field picture is a forward predicted frame, during coding with the reconstruction frames image of its former frame image as the reference frame; AVS encoder in the radio video sensor is that coding unit carries out the AVS coding to each two field picture in each image sets successively with the frame, and the two field picture of definition present encoding is a current frame image;
2. judge whether current frame image is the I two field picture, if each macro block that then travels through all intra prediction modes and be in the current frame image is selected best separately predictive mode, execution in step then
Otherwise, continue to carry out;
3. calculate the available code power consumption of current frame image, be designated as Pow,
Wherein, V is the supply power voltage of the cmos circuit in the radio video sensor, f
CLKClock frequency for the cmos circuit in the radio video sensor;
4. calculate current frame image and its reference frame absolute error and, be designated as SAD,
Wherein, Height represents the height of current frame image, and Width represents the width of current frame image, f
n(i, j) coordinate is (i, the pixel value of pixel j), f in the expression current frame image
N-1(i, j) coordinate is that (i, the pixel value of pixel j), n≤N, N represent the frame number of the two field picture that comprises in the vision signal in the expression reference frame;
5. absolute error and the SAD according to current frame image and its reference frame lists the AVS coding power consumption model:
Wherein, N
16 * 16And N
LeftSatisfy constraints:
D represents the coding distortion of current frame image, and M represents the number of the macro block that current frame image comprises,
N
16 * 16Adopt the number of the macro block of interframe 16 * 16 predictive modes coding in the expression current frame image, N
LeftAdopt the number of the macro block of all the other the inter-frame forecast mode codings except that Skip pattern and interframe 16 * 16 predictive modes in the expression current frame image, γ is an AVS coding power consumption model parameter, C
1The computation complexity of expression interframe 16 * 16 predictive modes, C
2The computation complexity of all the other inter-frame forecast modes of expression except that Skip pattern and interframe 16 * 16 predictive modes, C
3The computation complexity of entropy coding in the expression AVS encoder, all the other inter-frame forecast modes comprise interframe 16 * 8 predictive modes, interframe 8 * 16 predictive modes and interframe 8 * 8 predictive modes;
6. the parameter N in the cancellation AVS coding power consumption model
Left, obtain the coding distortion D of current frame image,
The value of calculating the coding distortion D make current frame image then adopts the number N of the macro block of interframe 16 * 16 predictive modes coding in hour current frame image
16 * 16
7. according to the number N of the macro block that adopts interframe 16 * 16 predictive modes coding in the current frame image
16 * 16, the number of the macro block of all the other inter-frame forecast modes codings of employing except that Skip pattern and interframe 16 * 16 predictive modes in the calculating current frame image
8. according to the number N of the macro block that adopts interframe 16 * 16 predictive modes coding in the current frame image
16 * 16Number N with the macro block that adopts all the other the inter-frame forecast mode codings except that Skip pattern and interframe 16 * 16 predictive modes in the current frame image
Left, calculate the number that adopts the macro block of Skip pattern-coding in the current frame image, be designated as N
Skip, N
Skip=M-N
16 * 16-N
Left
9. k the macro block that defines in the current frame image is current macro, calculate in the reference frame of current macro and current frame image with the absolute error of the corresponding macro block of current macro and, be designated as SAD
k,
f
kCoordinate is the pixel value of the pixel of (i ', j ') in (i ', j ') expression current macro, f
K_cIn the reference frame of (i ', j ') expression current frame image with the corresponding macro block of current macro in coordinate be the pixel value of the pixel of (i ', j '), 1≤k≤M;
10. make k '=k+1, k=k ' returns execution in step then and 9. the next macro block in the current frame image is handled, and all macro blocks dispose in current frame image;
By from small to large order in the reference frame of each macro block in the current frame image and current frame image with each macro block separately corresponding macro block absolute error and sort, constitute absolute error and sequence, be designated as { SAD
1, SAD
2..., SAD
p... SAD
M, from absolute error and sequence { SAD
1, SAD
2..., SAD
p... SAD
MThe middle N that takes out
SkipIndividual element
With (N
16 * 16+ N
Skip) individual element
Wherein, SAD
pExpression absolute error and sequence { SAD
1, SAD
2..., SAD
p... SAD
MIn p element;
Obtain in the current frame image absolute error of corresponding macro block in the reference frame with current frame image and smaller or equal to absolute error and sequence { SAD
1, SAD
2..., SAD
p... SAD
MIn N
SkipIndividual element
All macro blocks, the predictive mode of the best that these macro blocks need be adopted is defined as the Skip pattern, the sets definition that will be made up of these macro blocks is first set of macroblocks, the number of the macro block that first set of macroblocks comprises is N
Skip, the sets definition that will be made up of other macro blocks all macro blocks in first set of macroblocks in the current frame image is the residue set of macroblocks, the number of the macro block that the residue set of macroblocks comprises is M-N
Skip
Obtain in the residue set of macroblocks with the reference frame of current frame image in corresponding macro block absolute error and smaller or equal to absolute error and sequence { SAD
1, SAD
2..., SAD
p... SAD
MIn (N
16 * 16+ N
Skip) individual element
All macro blocks, the sets definition that these macro blocks are formed is second set of macroblocks, the number of the macro block that second set of macroblocks comprises is N
16 * 16, the sets definition that will be made up of other macro blocks all macro blocks in second set of macroblocks in the residue set of macroblocks is the 3rd set of macroblocks, the number of the macro block that the 3rd set of macroblocks comprises is M-N
Skip-N
16 * 16
Calculate the actual motion vector of each macro block in second set of macroblocks, the predictive mode of the best that each macro block of having determined the actual motion vector in second set of macroblocks need be adopted is defined as interframe 16 * 16 predictive modes;
Lagrangian cost when calculating each macro block in the 3rd set of macroblocks and adopting interframe 16 * 8 predictive modes, interframe 8 * 16 predictive modes and interframe 8 * 8 predictive modes respectively is with the Lagrangian cost predictive mode of hour corresponding inter-frame forecast mode the best that need adopt separately as each macro block;
Utilize the AVS encoder in the radio video sensor, each macro block in the current frame image is encoded with best separately predictive mode, finish until the whole frame coding of current frame image; Whether the supply power voltage V that detects the cmos circuit in the radio video sensor then is greater than 0, if two field picture then that the next frame in the vision signal is to be encoded is as current frame image, and return execution in step 2., otherwise the AVS encoder in the radio video sensor finishes coding.
The value of described AVS coding power consumption model parameter γ is 0.5.
Described step
Concrete steps be:
S the macro block that defines in second set of macroblocks is current macro, utilizes Lagrange cost function
Calculate the actual motion vector of current macro, wherein, MV '=(MV '
x, MV '
y), the actual motion vector of its expression current macro, J represents the Lagrangian cost of current macro,
f
sCoordinate is the pixel value of the pixel of (i ', j ') in (i ', j ') expression current macro, and c represents macro block identical with the current macro position in the reference frame of present frame, f
c(i '+MV
x, j '+MV
y) coordinate is the pixel value of pixel of the pixel distance MV ' of (i ', j ') among expression and the macro block c, λ
MOTIONBe the Lagrange's multiplier factor, the bit number that the actual motion vector MV ' coding of R (MV ') expression current macro may expend, 1≤s≤N
16 * 16 The predictive mode of the best that the current macro of determining the actual motion vector need be adopted is defined as interframe 16 * 16 predictive modes, makes s '=s+1, and s=s ' returns execution in step then
Next macro block in second set of macroblocks is handled, and all macro blocks dispose in second set of macroblocks.
Described step
Concrete steps be:
T the macro block that defines in the 3rd set of macroblocks is current macro, and current macro is divided into 2 16 * 8 sub-pieces, utilizes Lagrange cost function respectively
With
Calculate the actual motion vector of 2 16 * 8 sub-pieces, wherein,
It represents the actual motion vector of the 1st 16 * 8 sub-pieces,
It represents the actual motion vector of the 2nd 16 * 8 sub-pieces,
With
The Lagrangian cost of representing the 1st 16 * 8 sub-piece and the 2nd 16 * 8 sub-pieces respectively,
f
T1(i ", j ") and f
T2(i "; j ") represent respectively in the 1st 16 * 8 sub-piece and the 2nd 16 * 8 sub-piece coordinate for (i "; j ") the pixel value of pixel, l1 and l2 represent in the reference frame of present frame the 1st 16 * 8 sub-piece and the 2nd 16 * 8 identical sub-piece in sub-piece position with current macro respectively
The expression with sub-piece l1 in coordinate be (i ", the pixel distance of j ")
The pixel value of pixel,
The expression with sub-piece l2 in coordinate be (i ", the pixel distance of j ")
The pixel value of pixel, λ
MOTIONBe the Lagrange's multiplier factor,
The actual motion vector of representing the 1st 16 * 8 sub-pieces
The bit number that coding may expend,
The actual motion vector of representing the 2nd 16 * 8 sub-pieces
The bit number that coding may expend, BW=16, BH=8,1≤t≤M-N
Skip-N
16 * 16
Lagrangian cost according to the 1st 16 * 8 sub-pieces of current macro
Lagrangian cost with the 2nd 16 * 8 sub-pieces
Lagrangian cost when calculating current macro employing interframe 16 * 8 predictive modes is designated as J
16 * 8,
Current macro is divided into 28 * 16 sub-pieces, with step
With
Lagrangian cost when identical mode is calculated current macro employing interframe 8 * 16 predictive modes is designated as J
8 * 16
Current macro is divided into 48 * 8 sub-pieces, with step
With
Lagrangian cost when identical mode is calculated current macro employing interframe 8 * 8 predictive modes is designated as J
8 * 8
Lagrangian cost J when relatively current macro adopts interframe 16 * 8 predictive modes
16 * 8, the Lagrangian cost J when adopting interframe 8 * 16 predictive modes
8 * 16, the Lagrangian cost J when adopting interframe 8 * 8 predictive modes
8 * 8Size, with the Lagrangian cost predictive mode of hour corresponding inter-frame forecast mode the best that need adopt as current macro;
Make t '=t+1, t=t ' returns execution in step then
Next macro block in the 3rd set of macroblocks is handled, and all macro blocks dispose in the 3rd set of macroblocks.
The described Lagrange's multiplier factor
QP is the quantization parameter of the AVS encoder in the radio video sensor.
Compared with prior art, the invention has the advantages that by the I two field picture directly being carried out the AVS coding, and to the P two field picture, predictive mode is divided into the Skip pattern, interframe 16 * 16 predictive modes and all the other inter-frame forecast modes (comprising: interframe 16 * 8 predictive modes, interframe 8 * 16 predictive modes and interframe 8 * 8 predictive modes) three classes, according to listed radio video sensor power consumption model, when known target bit rate R and available code power consumption Pow, the macro block of determining to adopt the Skip pattern and adopting interframe 16 * 16 predictive modes coding, then utilize Lagrange cost function in all the other inter-frame forecast modes except that Skip pattern and interframe 16 * 16 predictive modes, to select best predictive mode to remaining macro block, guaranteed the low computational complexity and the validity of power consumption model; When the inventive method constantly reduces in the available code power consumption, can be in real time regulate in the two field picture in the vision signal and adopt the macro block number of Skip pattern, interframe 16 * 16 predictive modes and other coded in inter mode, thereby make the quality of radio video sensor video image under the situation of energy constraint keep optimum according to listed power consumption model; In addition, the inventive method has solved the contradiction between AVS coding power consumption, code check and the video quality of radio video sensor preferably, and computation complexity is low, and operand is little, can be widely used in the high wireless video sensor network that requires in real time.
Description of drawings
Fig. 1 is the FB(flow block) of the inventive method;
The structural representation of the vision signal that Fig. 2 obtains for radio video sensor;
Fig. 3 accounts for the ratio X of total macro block respectively for the macro block that adopts the inventive method to regulate to adopt the Skip pattern-coding, the macro block that adopts the macro block of interframe 16 * 16 predictive modes coding and adopt all the other inter-frame forecast modes to encode under different available code power consumption Pow, the result of Y and Z (
Cycle tests is " Monitor ", target bit rate R=0.5bpp).
Embodiment
Embodiment describes in further detail the present invention below in conjunction with accompanying drawing.
The AVS coding power consumption model optimization method of the radio video sensor that the present invention proposes, its FB(flow block) as shown in Figure 1, it specifically may further comprise the steps:
1. radio video sensor obtains vision signal from actual scene, the initial target bit rate of the current vision signal of obtaining of setting wireless video sensor, be designated as R, wherein, the structure of vision signal as shown in Figure 2, it comprises a plurality of image sets, each image sets mainly is made up of an I two field picture and at least one P two field picture, do not comprise the B two field picture in the image sets, start frame image when the I two field picture is encoded for each image sets, the P two field picture is a forward predicted frame, during coding with the reconstruction frames image of its former frame image as the reference frame; AVS encoder in the radio video sensor is that coding unit carries out the AVS coding to each two field picture in each image sets successively with the frame, and the two field picture of definition present encoding is a current frame image.
2. judge whether current frame image is the I two field picture, if each macro block that then travels through all intra prediction modes and be in the current frame image is selected best separately predictive mode, execution in step then
Otherwise, continue to carry out.
3. calculate the available code power consumption of current frame image, be designated as Pow,
Wherein, V is the supply power voltage of the cmos circuit in the radio video sensor, f
CLKClock frequency for the cmos circuit in the radio video sensor.
4. calculate current frame image and its reference frame absolute error and, be designated as SAD (Sum of AbsoluteDifference),
Wherein, Height represents the height of current frame image, and Width represents the width of current frame image, f
n(i, j) coordinate is (i, the pixel value of pixel j), f in the expression current frame image
N-1(i, j) coordinate is that (i, the pixel value of pixel j), n≤N, N represent the frame number of the two field picture that comprises in the vision signal in the expression reference frame.
5. absolute error and the SAD according to current frame image and its reference frame lists the AVS coding power consumption model:
Wherein, N
16 * 16And N
LeftSatisfy constraints:
D represents the coding distortion of current frame image, and M represents the number of the macro block that current frame image comprises,
For example every two field picture comprises 99 macro blocks, N in the vision signal of QCIF form (176 * 144)
16 * 16Adopt the number of the macro block of interframe 16 * 16 predictive modes coding in the expression current frame image, N
LeftAdopt the number of the macro block of all the other the inter-frame forecast mode codings except that Skip pattern and interframe 16 * 16 predictive modes in the expression current frame image, γ is an AVS coding power consumption model parameter, C
1The computation complexity of expression interframe 16 * 16 predictive modes, C
2The computation complexity of all the other inter-frame forecast modes of expression except that Skip pattern and interframe 16 * 16 predictive modes, C
3The computation complexity of entropy coding in the expression AVS encoder, all the other inter-frame forecast modes comprise interframe 16 * 8 predictive modes, interframe 8 * 16 predictive modes and interframe 8 * 8 predictive modes.
In this specific embodiment, the value of AVS coding power consumption model parameter γ is 0.5.
6. the parameter N in the cancellation AVS coding power consumption model
Left, obtain the coding distortion D of current frame image,
The value of calculating the coding distortion D make current frame image then adopts the number N of the macro block of interframe 16 * 16 predictive modes coding in hour current frame image
16 * 16
7. according to the number N of the macro block that adopts interframe 16 * 16 predictive modes coding in the current frame image
16 * 16, the number N of the macro block of all the other inter-frame forecast modes codings of employing except that Skip pattern and interframe 16 * 16 predictive modes in the calculating current frame image
Left,
8. according to the number N of the macro block that adopts interframe 16 * 16 predictive modes coding in the current frame image
16 * 16Number N with the macro block that adopts all the other the inter-frame forecast mode codings except that Skip pattern and interframe 16 * 16 predictive modes in the current frame image
Left, calculate the number that adopts the macro block of Skip pattern-coding in the current frame image, be designated as N
Skip, N
Skip=M-N
16 * 16-N
Left
9. k the macro block that defines in the current frame image is current macro, calculate in the reference frame of current macro and current frame image with the absolute error of the corresponding macro block of current macro and, be designated as SAD
k,
f
kCoordinate is the pixel value of the pixel of (i ', j ') in (i ', j ') expression current macro, f
K_cIn the reference frame of (i ', j ') expression current frame image with the corresponding macro block of current macro in coordinate be the pixel value of the pixel of (i ', j '), 1≤k≤M.
10. make k '=k+1, k=k ' returns execution in step then and 9. the next macro block in the current frame image is handled, and all macro blocks dispose in current frame image.
By from small to large order in the reference frame of each macro block in the current frame image and current frame image with each macro block separately corresponding macro block absolute error and sort, constitute absolute error and sequence, be designated as { SAD
1, SAD
2..., SAD
p... SAD
M, from absolute error and sequence { SAD
1, SAD
2..., SAD
p... SAD
MThe middle N that takes out
SkipIndividual element
With (N
16 * 16+ N
Skip) individual element
Wherein, SAD
pExpression absolute error and sequence { SAD
1, SAD
2..., SAD
p... SAD
MIn p element.
Obtain in the current frame image absolute error of corresponding macro block in the reference frame with current frame image and smaller or equal to absolute error and sequence { SAD
1, SAD
2..., SAD
p... SAD
MIn N
SkipIndividual element
All macro blocks, the predictive mode of the best that these macro blocks need be adopted is defined as the Skip pattern, the sets definition that will be made up of these macro blocks is first set of macroblocks, the number of the macro block that first set of macroblocks comprises is N
Skip, the sets definition that will be made up of other macro blocks all macro blocks in first set of macroblocks in the current frame image is the residue set of macroblocks, the number of the macro block that the residue set of macroblocks comprises is M-N
Skip
Obtain in the residue set of macroblocks with the reference frame of current frame image in corresponding macro block absolute error and smaller or equal to absolute error and sequence { SAD
1, SAD
2..., SAD
p... SAD
MIn (N
16 * 16+ N
Skip) individual element
All macro blocks, the sets definition that these macro blocks are formed is second set of macroblocks, the number of the macro block that second set of macroblocks comprises is N
16 * 16, the sets definition that will be made up of other macro blocks all macro blocks in second set of macroblocks in the residue set of macroblocks is the 3rd set of macroblocks, the number of the macro block that the 3rd set of macroblocks comprises is M-N
Skip-N
16 * 16
Calculate the actual motion vector of each macro block in second set of macroblocks, the predictive mode of the best that each macro block of having determined the actual motion vector in second set of macroblocks need be adopted is defined as interframe 16 * 16 predictive modes.Concrete steps are:
S the macro block that defines in second set of macroblocks is current macro, utilizes Lagrange cost function
Calculate the actual motion vector of current macro, wherein, MV '=(MV '
x, MV '
y), the actual motion vector of its expression current macro, J represents the Lagrangian cost of current macro,
f
sCoordinate is the pixel value of the pixel of (i ', j ') in (i ', j ') expression current macro, and c represents macro block identical with the current macro position in the reference frame of present frame, f
c(i '+MV
x, j '+MV
y) coordinate is the pixel value of pixel of the pixel distance MV ' of (i ', j ') among expression and the macro block c, λ
MOTIONBe the Lagrange's multiplier factor, the bit number that the actual motion vector MV ' coding of R (MV ') expression current macro may expend, 1≤s≤N
16 * 16 The predictive mode of the best that the current macro of determining the actual motion vector need be adopted is defined as interframe 16 * 16 predictive modes, makes s '=s+1, and s=s ' returns execution in step then
Next macro block in second set of macroblocks is handled, and all macro blocks dispose in second set of macroblocks.
Lagrangian cost when calculating each macro block in the 3rd set of macroblocks and adopting interframe 16 * 8 predictive modes, interframe 8 * 16 predictive modes and interframe 8 * 8 predictive modes respectively is with the Lagrangian cost predictive mode of hour corresponding inter-frame forecast mode the best that need adopt separately as each macro block.Concrete steps are:
T the macro block that defines in the 3rd set of macroblocks is current macro, and current macro is divided into 2 16 * 8 sub-pieces, utilizes Lagrange cost function respectively
With
Calculate the actual motion vector of 2 16 * 8 sub-pieces, wherein,
It represents the actual motion vector of the 1st 16 * 8 sub-pieces,
It represents the actual motion vector of the 2nd 16 * 8 sub-pieces,
With
The Lagrangian cost of representing the 1st 16 * 8 sub-piece and the 2nd 16 * 8 sub-pieces respectively,
f
T1(i ", j ") and f
T2(i "; j ") represent respectively in the 1st 16 * 8 sub-piece and the 2nd 16 * 8 sub-piece coordinate for (i "; j ") the pixel value of pixel, l1 and l2 represent in the reference frame of present frame the 1st 16 * 8 sub-piece and the 2nd 16 * 8 identical sub-piece in sub-piece position with current macro respectively
The expression with sub-piece l1 in coordinate be (i ", the pixel distance of j ")
The pixel value of pixel,
The expression with sub-piece l2 in coordinate be (i ", the pixel distance of j ")
The pixel value of pixel, λ
MOTIONBe the Lagrange's multiplier factor,
The actual motion vector of representing the 1st 16 * 8 sub-pieces
The bit number that coding may expend,
The actual motion vector of representing the 2nd 16 * 8 sub-pieces
The bit number that coding may expend, BW=16, BH=8,1≤t≤M-N
Skip-N
16 * 16
Lagrangian cost according to the 1st 16 * 8 sub-pieces of current macro
Lagrangian cost with the 2nd 16 * 8 sub-pieces
Lagrangian cost when calculating current macro employing interframe 16 * 8 predictive modes is designated as J
16 * 8,
Current macro is divided into 28 * 16 sub-pieces, with step
With
Lagrangian cost when identical mode is calculated current macro employing interframe 8 * 16 predictive modes is designated as J
8 * 16
Current macro is divided into 48 * 8 sub-pieces, with step
With
Lagrangian cost when identical mode is calculated current macro employing interframe 8 * 8 predictive modes is designated as J
8 * 8
Lagrangian cost J when relatively current macro adopts interframe 16 * 8 predictive modes
16 * 8, the Lagrangian cost J when adopting interframe 8 * 16 predictive modes
8 * 16, the Lagrangian cost J when adopting interframe 8 * 8 predictive modes
8 * 8Size, with the Lagrangian cost predictive mode of hour corresponding inter-frame forecast mode the best that need adopt as current macro.
Make t '=t+1, t=t ' returns execution in step then
Next macro block in the 3rd set of macroblocks is handled, and all macro blocks dispose in the 3rd set of macroblocks.
Utilize the AVS encoder in the radio video sensor, each macro block in the current frame image is encoded with best separately predictive mode, finish until the whole frame coding of current frame image; Whether the supply power voltage V that detects the cmos circuit in the radio video sensor then is greater than 0, if two field picture then that the next frame in the vision signal is to be encoded is as current frame image, and return execution in step 2., otherwise the AVS encoder in the radio video sensor finishes coding.
In this specific embodiment, the Lagrange's multiplier factor
QP is the quantization parameter of the AVS encoder in the radio video sensor.
Fig. 3 has provided the macro block that adopts the inventive method to regulate employing Skip pattern-coding under different available code power consumption Pow, the ratio X that macro block that adopts interframe 16 * 16 predictive modes coding and the macro block that adopts all the other inter-frame forecast modes (interframe 16 * 8 predictive modes, interframe 8 * 16 predictive modes and interframe 8 * 8 predictive modes) to encode account for total macro block respectively, the result of Y and Z, wherein
Cycle tests adopts " Monitor ", and target bit rate is R=0.5bpp.As can be seen from Figure 3 along with the continuous increase of available code power consumption Pow, the ratio that the macro block that the macro block that adopts interframe 16 * 16 predictive modes to encode accounts for the ratio of total macro block and adopts all the other inter-frame forecast modes to encode accounts for total macro block constantly increases, and the macro block that adopts the Skip pattern-coding accounts for the ratio of total macro block and constantly reduces, this be because: on the one hand, interframe 16 * 16 predictive modes and all the other inter-frame forecast modes need more power consumption than Skip pattern; On the other hand, when more power consumption was provided, in order to make the distortion minimum of the two field picture in the vision signal, more macro block will select more accurate interframe 16 * 16 predictive modes of prediction and all the other inter-frame forecast modes to encode.
Claims (5)
1. the AVS coding power consumption model optimization method of a radio video sensor is characterized in that may further comprise the steps:
1. the initial target bit rate of the current vision signal of obtaining of setting wireless video sensor, be designated as R, wherein, vision signal comprises a plurality of image sets, each image sets mainly is made up of an I two field picture and at least one P two field picture, the start frame image of I two field picture during for each image sets coding, the P two field picture is a forward predicted frame, during coding with the reconstruction frames image of its former frame image as the reference frame; AVS encoder in the radio video sensor is that coding unit carries out the AVS coding to each two field picture in each image sets successively with the frame, and the two field picture of definition present encoding is a current frame image;
2. judge whether current frame image is the I two field picture, if each macro block that then travels through all intra prediction modes and be in the current frame image is selected best separately predictive mode, execution in step then
Otherwise, continue to carry out;
3. calculate the available code power consumption of current frame image, be designated as Pow,
Wherein, V is the supply power voltage of the cmos circuit in the radio video sensor, f
CLKClock frequency for the cmos circuit in the radio video sensor;
4. calculate current frame image and its reference frame absolute error and, be designated as SAD,
Wherein, Height represents the height of current frame image, and Width represents the width of current frame image, f
n(i, j) coordinate is (i, the pixel value of pixel j), f in the expression current frame image
N-1(i, j) coordinate is that (i, the pixel value of pixel j), n≤N, N represent the frame number of the two field picture that comprises in the vision signal in the expression reference frame;
5. absolute error and the SAD according to current frame image and its reference frame lists the AVS coding power consumption model:
Wherein, N
16 * 16And N
LeftSatisfy constraints:
D represents the coding distortion of current frame image, and M represents the number of the macro block that current frame image comprises,
N
16 * 16Adopt the number of the macro block of interframe 16 * 16 predictive modes coding in the expression current frame image, N
LeftAdopt the number of the macro block of all the other the inter-frame forecast mode codings except that Skip pattern and interframe 16 * 16 predictive modes in the expression current frame image, γ is an AVS coding power consumption model parameter, C
1The computation complexity of expression interframe 16 * 16 predictive modes, C
2The computation complexity of all the other inter-frame forecast modes of expression except that Skip pattern and interframe 16 * 16 predictive modes, C
3The computation complexity of entropy coding in the expression AVS encoder, all the other inter-frame forecast modes comprise interframe 16 * 8 predictive modes, interframe 8 * 16 predictive modes and interframe 8 * 8 predictive modes;
6. the parameter N in the cancellation AVS coding power consumption model
Left, obtain the coding distortion D of current frame image,
, the value of calculating the coding distortion D make current frame image then adopts the number N of the macro block of interframe 16 * 16 predictive modes coding in hour current frame image
16 * 16
7. according to the number N of the macro block that adopts interframe 16 * 16 predictive modes coding in the current frame image
16 * 16, the number N of the macro block of all the other inter-frame forecast modes codings of employing except that Skip pattern and interframe 16 * 16 predictive modes in the calculating current frame image
Left,
8. according to the number N of the macro block that adopts interframe 16 * 16 predictive modes coding in the current frame image
16 * 16Number N with the macro block that adopts all the other the inter-frame forecast mode codings except that Skip pattern and interframe 16 * 16 predictive modes in the current frame image
Left, calculate the number that adopts the macro block of Skip pattern-coding in the current frame image, be designated as N
Skip, N
Skip=M-N
16 * 16-N
Left
9. k the macro block that defines in the current frame image is current macro, calculate in the reference frame of current macro and current frame image with the absolute error of the corresponding macro block of current macro and, be designated as SAD
k,
f
kCoordinate is the pixel value of the pixel of (i ', j ') in (i ', j ') expression current macro, f
K_cIn the reference frame of (i ', j ') expression current frame image with the corresponding macro block of current macro in coordinate be the pixel value of the pixel of (i ', j '), 1≤k≤M;
10. make k '=k+1, k=k ' returns execution in step then and 9. the next macro block in the current frame image is handled, and all macro blocks dispose in current frame image;
By from small to large order in the reference frame of each macro block in the current frame image and current frame image with each macro block separately corresponding macro block absolute error and sort, constitute absolute error and sequence, be designated as { SAD
1, SAD
2..., SAD
p... SAD
M, from absolute error and sequence { SAD
1, SAD
2..., SAD
p... SAD
MThe middle N that takes out
SkipIndividual element
With (N
16 * 16+ N
Skip) individual element
Wherein, SAD
pExpression absolute error and sequence { SAD
1, SAD
2..., SAD
p... SAD
MIn p element;
Obtain in the current frame image absolute error of corresponding macro block in the reference frame with current frame image and smaller or equal to absolute error and sequence { SAD
1, SAD
2..., SAD
p... SAD
MIn N
SkipIndividual element
All macro blocks, the predictive mode of the best that these macro blocks need be adopted is defined as the Skip pattern, the sets definition that will be made up of these macro blocks is first set of macroblocks, the number of the macro block that first set of macroblocks comprises is N
Skip, the sets definition that will be made up of other macro blocks all macro blocks in first set of macroblocks in the current frame image is the residue set of macroblocks, the number of the macro block that the residue set of macroblocks comprises is M-N
Skip
Obtain in the residue set of macroblocks with the reference frame of current frame image in corresponding macro block absolute error and smaller or equal to absolute error and sequence { SAD
1, SAD
2..., SAD
p... SAD
MIn (N
16 * 16+ N
Skip) individual element
All macro blocks, the sets definition that these macro blocks are formed is second set of macroblocks, the number of the macro block that second set of macroblocks comprises is N
16 * 16, the sets definition that will be made up of other macro blocks all macro blocks in second set of macroblocks in the residue set of macroblocks is the 3rd set of macroblocks, the number of the macro block that the 3rd set of macroblocks comprises is M-N
Skip-N
16 * 16
Calculate the actual motion vector of each macro block in second set of macroblocks, the predictive mode of the best that each macro block of having determined the actual motion vector in second set of macroblocks need be adopted is defined as interframe 16 * 16 predictive modes;
Lagrangian cost when calculating each macro block in the 3rd set of macroblocks and adopting interframe 16 * 8 predictive modes, interframe 8 * 16 predictive modes and interframe 8 * 8 predictive modes respectively is with the Lagrangian cost predictive mode of hour corresponding inter-frame forecast mode the best that need adopt separately as each macro block;
Utilize the AVS encoder in the radio video sensor, each macro block in the current frame image is encoded with best separately predictive mode, finish until the whole frame coding of current frame image; Whether the supply power voltage V that detects the cmos circuit in the radio video sensor then is greater than 0, if two field picture then that the next frame in the vision signal is to be encoded is as current frame image, and return execution in step 2., otherwise the AVS encoder in the radio video sensor finishes coding.
2. the AVS coding power consumption model optimization method of a kind of radio video sensor according to claim 1 is characterized in that the value of described AVS coding power consumption model parameter γ is 0.5.
3. the AVS coding power consumption model optimization method of a kind of radio video sensor according to claim 1 and 2 is characterized in that described step
Concrete steps be:
S the macro block that defines in second set of macroblocks is current macro, utilizes Lagrange cost function
Calculate the actual motion vector of current macro, wherein, MV '=(MV '
x, MV '
y), the actual motion vector of its expression current macro, J represents the Lagrangian cost of current macro,
f
sCoordinate is the pixel value of the pixel of (i ', j ') in (i ', j ') expression current macro, and c represents macro block identical with the current macro position in the reference frame of present frame, f
c(i '+MV
x, j '+MV
y) coordinate is the pixel value of pixel of the pixel distance MV ' of (i ', j ') among expression and the macro block c, λ
MOTIONBe the Lagrange's multiplier factor, the bit number that the actual motion vector MV ' coding of R (MV ') expression current macro may expend, 1≤s≤N
16 * 16 The predictive mode of the best that the current macro of determining the actual motion vector need be adopted is defined as interframe 16 * 16 predictive modes, makes s '=s+1, and s=s ' returns execution in step then
Next macro block in second set of macroblocks is handled, and all macro blocks dispose in second set of macroblocks.
4. the AVS coding power consumption model optimization method of a kind of radio video sensor according to claim 3 is characterized in that described step
Concrete steps be:
T the macro block that defines in the 3rd set of macroblocks is current macro, and current macro is divided into 2 16 * 8 sub-pieces, utilizes Lagrange cost function respectively
With
Calculate the actual motion vector of 2 16 * 8 sub-pieces, wherein,
It represents the actual motion vector of the 1st 16 * 8 sub-pieces,
It represents the actual motion vector of the 2nd 16 * 8 sub-pieces,
With
The Lagrangian cost of representing the 1st 16 * 8 sub-piece and the 2nd 16 * 8 sub-pieces respectively,
f
T1(i ", j ") and f
T2(i "; j ") represent respectively in the 1st 16 * 8 sub-piece and the 2nd 16 * 8 sub-piece coordinate for (i "; j ") the pixel value of pixel, l1 and l2 represent in the reference frame of present frame the 1st 16 * 8 sub-piece and the 2nd 16 * 8 identical sub-piece in sub-piece position with current macro respectively
The expression with sub-piece l1 in coordinate be (i ", the pixel distance of j ")
The pixel value of pixel,
The expression with sub-piece l2 in coordinate be (i ", the pixel distance of j ")
The pixel value of pixel, λ
MOTIONBe the Lagrange's multiplier factor,
The actual motion vector of representing the 1st 16 * 8 sub-pieces
The bit number that coding may expend,
The actual motion vector of representing the 2nd 16 * 8 sub-pieces
The bit number that coding may expend, BW=16, BH=8,1≤t≤M-N
Skip-N
16 * 16
Lagrangian cost according to the 1st 16 * 8 sub-pieces of current macro
Lagrangian cost with the 2nd 16 * 8 sub-pieces
Lagrangian cost when calculating current macro employing interframe 16 * 8 predictive modes is designated as J16 * 8,
Current macro is divided into 28 * 16 sub-pieces, with step
With
Lagrangian cost when identical mode is calculated current macro employing interframe 8 * 16 predictive modes is designated as J
8 * 16
Current macro is divided into 48 * 8 sub-pieces, with step
With
Lagrangian cost when identical mode is calculated current macro employing interframe 8 * 8 predictive modes is designated as J
8 * 8
Lagrangian cost J when relatively current macro adopts interframe 16 * 8 predictive modes
16 * 8, the Lagrangian cost J when adopting interframe 8 * 16 predictive modes
8 * 16, the Lagrangian cost J when adopting interframe 8 * 8 predictive modes
8 * 8Size, with the Lagrangian cost predictive mode of hour corresponding inter-frame forecast mode the best that need adopt as current macro;
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201010184247 CN101848394B (en) | 2010-05-25 | 2010-05-25 | Method for optimizing AVS coding power consumption model of radio video sensor |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN 201010184247 CN101848394B (en) | 2010-05-25 | 2010-05-25 | Method for optimizing AVS coding power consumption model of radio video sensor |
Publications (2)
Publication Number | Publication Date |
---|---|
CN101848394A true CN101848394A (en) | 2010-09-29 |
CN101848394B CN101848394B (en) | 2012-01-18 |
Family
ID=42772823
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN 201010184247 Expired - Fee Related CN101848394B (en) | 2010-05-25 | 2010-05-25 | Method for optimizing AVS coding power consumption model of radio video sensor |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN101848394B (en) |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105516722A (en) * | 2010-09-30 | 2016-04-20 | 三菱电机株式会社 | Moving image encoding device, moving image decoding device, moving image coding method, and moving image decoding method |
WO2023098636A1 (en) * | 2021-11-30 | 2023-06-08 | Huawei Technologies Co., Ltd. | Method, device, and medium for adaptive inference in compressed video domain |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2005004491A1 (en) * | 2003-06-25 | 2005-01-13 | Thomson Licensing S.A. | Fast mode-decision encoding for interframes |
CN101022555A (en) * | 2007-02-12 | 2007-08-22 | 清华大学 | Interframe predictive coding mode quick selecting method |
CN101141647A (en) * | 2007-08-24 | 2008-03-12 | 上海广电(集团)有限公司中央研究院 | AVS video coding based fast intraframe predicting mode selecting method |
CN101159873A (en) * | 2007-11-16 | 2008-04-09 | 中国科学院计算技术研究所 | Inter-frame mode selecting method |
CN101448159A (en) * | 2009-01-08 | 2009-06-03 | 北京航空航天大学 | Rapid interframe mode selection method based on rate-distortion cost and mode frequency |
-
2010
- 2010-05-25 CN CN 201010184247 patent/CN101848394B/en not_active Expired - Fee Related
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2005004491A1 (en) * | 2003-06-25 | 2005-01-13 | Thomson Licensing S.A. | Fast mode-decision encoding for interframes |
CN101022555A (en) * | 2007-02-12 | 2007-08-22 | 清华大学 | Interframe predictive coding mode quick selecting method |
CN101141647A (en) * | 2007-08-24 | 2008-03-12 | 上海广电(集团)有限公司中央研究院 | AVS video coding based fast intraframe predicting mode selecting method |
CN101159873A (en) * | 2007-11-16 | 2008-04-09 | 中国科学院计算技术研究所 | Inter-frame mode selecting method |
CN101448159A (en) * | 2009-01-08 | 2009-06-03 | 北京航空航天大学 | Rapid interframe mode selection method based on rate-distortion cost and mode frequency |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105516722A (en) * | 2010-09-30 | 2016-04-20 | 三菱电机株式会社 | Moving image encoding device, moving image decoding device, moving image coding method, and moving image decoding method |
CN105516722B (en) * | 2010-09-30 | 2018-12-07 | 三菱电机株式会社 | Dynamic image encoding device and method, moving image decoding apparatus and method |
WO2023098636A1 (en) * | 2021-11-30 | 2023-06-08 | Huawei Technologies Co., Ltd. | Method, device, and medium for adaptive inference in compressed video domain |
Also Published As
Publication number | Publication date |
---|---|
CN101848394B (en) | 2012-01-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN101815218B (en) | Method for coding quick movement estimation video based on macro block characteristics | |
CN103188496B (en) | Based on the method for coding quick movement estimation video of motion vector distribution prediction | |
CN103327325A (en) | Intra-frame prediction mode rapid self-adaptation selection method based on HEVC standard | |
CN101835042B (en) | Wyner-Ziv video coding system controlled on the basis of non feedback speed rate and method | |
CN101022555B (en) | Interframe predictive coding mode quick selecting method | |
CN102740077B (en) | H.264/AVC standard-based intra-frame prediction mode selection method | |
CN105049850A (en) | HEVC (High Efficiency Video Coding) code rate control method based on region-of-interest | |
CN101860748A (en) | Side information generating system and method based on distribution type video encoding | |
CN102209243A (en) | Depth map intra prediction method based on linear model | |
CN102630013A (en) | Bit rate control video compression method and device on basis of scene switching | |
CN101014129B (en) | Video data compression method | |
CN103634606A (en) | Video encoding method and apparatus | |
CN103533359A (en) | H.264 code rate control method | |
CN102196272B (en) | P frame encoding method and device | |
CN101562750B (en) | Device and method for fast selecting video coding mode | |
CN103957410A (en) | I-frame code rate control method based on residual frequency domain complexity | |
CN102148990B (en) | Device and method for predicting motion vector | |
CN104702959B (en) | A kind of intra-frame prediction method and system of Video coding | |
CN102572428B (en) | Side information estimating method oriented to distributed coding and decoding of multimedia sensor network | |
CN101715124B (en) | Single-input and multi-output video encoding system and video encoding method | |
CN103384327A (en) | AVS fast mode selection algorithm based on adaptive threshold | |
CN104754337A (en) | Video encoding method | |
CN101883275B (en) | Video coding method | |
CN104113761A (en) | Code rate control method for video encoding and encoder | |
CN102420987A (en) | Self-adaption bit distribution method based on code rate control of hierarchical B frame structure |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
C06 | Publication | ||
PB01 | Publication | ||
C10 | Entry into substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
C14 | Grant of patent or utility model | ||
GR01 | Patent grant | ||
CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20120118 Termination date: 20150525 |
|
EXPY | Termination of patent right or utility model |