CN109474825A - A kind of compressing pulse trains method and system - Google Patents
A kind of compressing pulse trains method and system Download PDFInfo
- Publication number
- CN109474825A CN109474825A CN201811217843.1A CN201811217843A CN109474825A CN 109474825 A CN109474825 A CN 109474825A CN 201811217843 A CN201811217843 A CN 201811217843A CN 109474825 A CN109474825 A CN 109474825A
- Authority
- CN
- China
- Prior art keywords
- block
- information
- sequence
- sub
- grey level
- 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
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
- H04N19/13—Adaptive entropy coding, e.g. adaptive variable length coding [AVLC] or context adaptive binary arithmetic coding [CABAC]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/169—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding
- H04N19/17—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object
- H04N19/176—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the coding unit, i.e. the structural portion or semantic portion of the video signal being the object or the subject of the adaptive coding the unit being an image region, e.g. an object the region being a block, e.g. a macroblock
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
- H04N19/61—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding in combination with predictive coding
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/85—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using pre-processing or post-processing specially adapted for video compression
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/90—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using coding techniques not provided for in groups H04N19/10-H04N19/85, e.g. fractals
- H04N19/91—Entropy coding, e.g. variable length coding [VLC] or arithmetic coding
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The present invention mainly proposes a kind of compressing pulse trains method and system, comprising: original burst signal is converted to sequence of grey level；Block division is carried out to the sequence of grey level, obtains several sub-blocks；Each sub-block is predicted to obtain predicted pixel values, calculates the residual error of true value and predicted pixel values；Transform and quantization operation is carried out to the information of the residual error；Information, predictive information and the residual information that described piece divides further are compressed by entropy coding, and are stored as binary code stream.This method takes full advantage of correlation of the pulse sequence signal on room and time, to effectively compress pulse train.The experimental results showed that this method can be significantly compression pulse train, can effectively apply in the compression, transmission, storage system of actual pulse train content.
Description
Technical field
The invention belongs to digital processing fields, and in particular to one kind is by dynamic visual sensor (DVS, Dynamic
Vision Sensors) record pulse train compression method and system.
Background technique
Dynamic visual sensor is obtained it as nerve signal by imitating retina to sense and encode the world
Visual information, so dynamic visual sensor is a kind of mind of autokinetic movement control that is promising, can be used for mobile robot
Through form vision sensor.Although researcher perceives environment using various sensors, such as based on the camera of frame, structure
Optical sensor and stereoscopic camera etc., but still there are many limitation and defects.As a promising solution, dynamic vision
Feel sensor by simulating retina, and the arteries and veins that response is generated due to the Pixel-level brightness change in scene caused by moving
Punching.Compared with traditional video camera based on frame, especially for moving field, DVS is in data rate, speed and dynamic range side
Face has very big advantage.In addition, impulsive neural networks (SNN, Spiking can be transferred directly to by the pulse that DVS is generated
Neural Network), it is used for visual processes and motion control.
With the development of video technique, there are higher need to the dynamic range of video and temporal resolution in many scenes
It asks, the advantage of dynamic vision perceptron is emerged from these scenes.The frame per second of traditional camera is generally tens, higher frame
Rate often greatlys improve technology and production cost.And dynamic visual sensor is the pulse letter of record reflection motion information
Number, frame per second can achieve frames up to ten thousand, have broad application prospects under the high speeds motion photography such as automatic Pilot.
Dynamic visual sensor is a kind of visual sensor of novel class retina.In dynamic visual sensor, often
A pixel by responding brightness change and being encoded with generating asynchronous indie incident, the certain journey of flow of event generated
The redundancy in the consecutive image of traditional cameras output in time domain is eliminated on degree；And it has high temporal resolution,
Supper-fast movement can be captured；It, i.e., can work well at daytime and night in addition, it has a very high dynamic range
Make.So dynamic visual sensor can also be used in monitoring system.
The pulse signal that DVS is generated generally indicates (AER, Address Event Representation) with address events
Form storage, each data be by event address (position of respective pixel) and event property (brighten or
It is dimmed) etc. composition.Since DVS frame per second is high, data volume is also very big, needs to occupy very big transmission bandwidth and memory space,
It is objectively too high to the requirement of software and hardware.In addition to this, existing DVS pulse train processing method, can not be integrated into newest
In video encoding and decoding standard, the operations such as subsequent compression can not be carried out.
Summary of the invention
For the deficiency for solving existing compressing pulse trains technology, the present invention, which provides one kind, can effectively compress AER pulse
The method of sequence.By the way that pulse signal is synthesized gray level image, and lossy coding is being carried out later, can greatly reduce transmission
Bandwidth and storage cost.
Specifically, according to an aspect of the present invention, additionally providing a kind of compressing pulse trains method, comprising:
Original burst signal is converted into sequence of grey level；
Block division is carried out to the sequence of grey level, obtains several sub-blocks；
Each sub-block is predicted to obtain predicted pixel values, calculates the residual error of true value and predicted pixel values；
Transform and quantization operation is carried out to the information of the residual error；
Information, predictive information and the residual information that described piece divides further are compressed by entropy coding, and are stored as two
System code stream.
It is preferably, described that original burst signal is converted into sequence of grey level, comprising:
By between the original address added-time, the pulse signal of polar format storage be converted to the pulse image of 1 bit-depth；
Every continuous several above-mentioned pulse images of frame are synthesized to the single channel image of a frame multiple bit-depth.
Preferably, described piece of division include:
The sequence of grey level is divided into the cube of multiple full resolutions, the space point of each cube in time
Resolution is identical as sequence of grey level；
Each full resolution cube is spatially divided into smaller sub-block.
Preferably, the block prediction being predicted as in airspace, the predicted pixel values are by same full resolution cube
Boundary pixel in encoded adjacent sub-blocks generates.
Preferably, the block prediction being predicted as in time domain, uses the encoded sub-block in neighbouring full resolution cube
Predict the pixel value in present encoding sub-block.
Preferably, the entropy coding is carried out adaptive by predictive information and residual information of the context model to each sub-block
Answer binary arithmetic coding.
Preferably, the binary arithmetic coding saves coding based on the mode of recurrence interval division in a recursive process
Section and interval limit；Using adaptive probabilistic model, after current syntax element binarization, each binary system text
Part carries out arithmetic coding according to its probabilistic model parameter.
According to another aspect of the present invention, a kind of compressing pulse trains system is also disclosed, comprising:
Conversion module, for original burst signal to be converted to sequence of grey level；
Block division module obtains several sub-blocks for carrying out block division to the sequence of grey level；
Prediction module obtains predicted pixel values for being predicted each sub-block, calculates true value and prediction picture
The residual error of element value；
Transform and quantization module carries out transform and quantization operation for the information to the residual error.
Entropy code module, the information, predictive information and residual information for being divided described piece by entropy coding are further
Compression, and it is stored as binary code stream.
The present invention has the advantages that transmission bandwidth and storage cost can be simply and effectively reduced, complexity is lower, can be with
It effectively applies in the relevant compression of pulse train, transmission, storage system.
Detailed description of the invention
By reading the following detailed description of the preferred embodiment, various other advantages and benefits are common for this field
Technical staff will become clear.The drawings are only for the purpose of illustrating a preferred embodiment, and is not considered as to the present invention
Limitation.And throughout the drawings, the same reference numbers will be used to refer to the same parts.In the accompanying drawings:
Attached drawing 1 is a kind of pulse train coding method flow chart provided by the invention；
Attached drawing 2 is display diagram of the pulse signal provided by the invention to image area；
Attached drawing 3 is the schematic diagram that pulse signal provided by the invention synthesizes 8 bit-depth grayscale images；
Attached drawing 4 is provided by the invention piece of division schematic diagram；
Attached drawing 5 is lossy compression coded system structural schematic diagram provided by the invention.
Specific embodiment
The illustrative embodiments of the disclosure are more fully described below with reference to accompanying drawings.Although showing this public affairs in attached drawing
The illustrative embodiments opened, it being understood, however, that may be realized in various forms the disclosure without the reality that should be illustrated here
The mode of applying is limited.It is to be able to thoroughly understand the disclosure on the contrary, providing these embodiments, and can be by this public affairs
The range opened is fully disclosed to those skilled in the art.
The concept of bit-depth (Bit Depth) is widely used in digital video-audio field, mainly uses in digital picture
The bit number used in the single color component of expression pixel, but usually referred to as color depth (Color Depth) or quantization is deep
Degree, then for expressing bit number used in sound sample value in digital audio.
In digital image arts, bit-depth determines the number of colours that digital picture can be expressed, thus determines color
The levels of precision of expression.For example, 1 bit can express 2 kinds of colors (claiming monochrome, usually black and white), 2 bits can express 4 kinds of colors,
4 bits can express 16 kinds of colors, and 8 bits can express 256 kinds of colors, and so on.It is deep in the color for specifically describing piece image
When spending, " bit number of each pixel " (bits per pixel, abbreviation bpp) is usually used to express, such as digital movie is adopted
With 36bpp, i.e. 36 bits/pixels.Using certain bit-depth, different gray scales (brightness) can be expressed.But works as and be used for table
Up to each color bit number be less than 8 when, apparent striated or patch shape can be presented in the color of image, and this phenomenon is known as color
Adjust separation (Posterization).Human eye can only distinguish about 10,000,000 kinds of different colors, therefore, if in order to
Viewing, the image of 24bpp just can satisfy needs under normal circumstances, just be shown with the bit-depth storage image higher than 24bpp
It is extra to obtain.However, still useful higher than the image of 24bpp, it can keep higher quality in digital post-processing.
According to wanting for DCI digital movie system specification (DCSS, Digital Cinema System Specification)
It asks, the bit-depth of each color component is 12 bits in digital film images, and each pixel is made of three color components, because
And the bit-depth of each pixel is 36 bits, i.e. 36bpp.Digital cinema sound sampling frequency is 48kHz/ sound channel or 96kHz/
Sound channel, each sampling value carry out the quantization that depth is 24 bits.
Firstly, the original pulse signal with address format record is converted to the ash of n-bit depth by method of the invention
Spend graphic sequence.The value of n can flexibly be chosen in practical application, and conventional value is 8 or 10.These pulses are believed under normal conditions
Number there is stronger correlation on spatially and temporally, therefore the present invention is carried out by removing these correlations in the next steps
Lossy compression.
For lossy compression process, the present invention is separately encoded after obtained sequence of grey level is carried out block division, Zhi Houtong
It crosses intra prediction or inter-prediction predicts current block, carry out transform and quantization is carried out to the difference of true value and predicted value
Operation, then encodes, and eliminates the correlation on spatially and temporally and improves compression efficiency, finally by block division information, prediction letter
The adaptive entropy coding of information needed merga pass such as breath and residual error data further compresses, and is finally stored as binary code stream.
Transformation is so that the energy of residual matrix is more concentrated, then further increases compression efficiency by specific quantization matrix.Quantifying
Due to the source of loss during having ignored partial information therefore being irreversible and lossy coding in operation.In order to control
The size of distortion carrys out adjustment quantization matrix with a quantization parameter (QP, quantization parameter) here.
The lossy compression encryption algorithm of 1 pulse train of embodiment
Specifically, as shown in Figure 1, according to an aspect of the present invention, the invention proposes a kind of compressing pulse trains sides
Method includes the following steps:
S1, pulse signal is converted into sequence of grey level
Each pulse signal is indicated that x and y respectively indicate pulse letter by the four-tuple (x, y, t, p) of a fixed number of bits
Abscissa and ordinate number on the image, p indicate that the polarity of pulse signal (is indicated, value is 0 or 1), when t is indicated with 1bit
Between axis.Pulse signal is gone to image area by the present invention, then every continuous n frame is converted to the image of a n-bit depth.
Firstly, the time domain of original burst signal and spatial information (si) are converted to the pulse image of 1 bit-depth by this method.It is right
In the pulse train { (x that length is n1, y1, t1, p1), (x2, y2, t2, p2) ..., (xn, yn, tn, pn), according to it in time shaft
On length generate tn-t1+ 1 frame, the initial value of all pixels point intensity are 0, the pixel depth of each frame be 1 bit (i.e.
Value is 0 or 1).Later for each pulse signal (xi,yi,ti,pi), by ti-t1+ 1 frame (xi,yi) at pixel set
It is set to 1, so that pulse signal be indicated in the form of images.
And then every continuous several above-mentioned pulse images of frame are synthesized to the single channel image of a frame n-bit depth, more
Conventional value is that take n be 8 or 10, i.e., the pulse image of 1 bit-depth of 8 frames or 10 frame is synthesized 8 bits or 10
The grayscale image of bit-depth needs that the other values other than 8 and 10 can also be chosen depending on concrete condition.By z frame to z+
The method that n-1 frame synthesizes a frame n-bit depth grayscale image can be expressed with following formula:
Wherein aU, v, z+iIndicate that coordinate is the value at (u, v), b on z+i frame pulse imageU, vThen indicate this n frame pulse figure
Value of the grayscale image of picture synthesis at (u, v).
S2, block divide
Cube of the invention first that sequence of grey level obtained in the first step is divided into multiple full resolutions in time
The spatial resolution of body, each cube is identical as sequence of grey level.Each full resolution cube is spatially drawn later
It is divided into smaller sub-block to carry out subsequent operation.
S3, prediction
When encoding each sub-block, the present invention is predicted to it, then encodes predictive information and true value and prediction
The residual error of value.The range that match block is searched for when according to prediction, can be divided into both of which for prediction.
The first is the block prediction in airspace, and in such a mode, the predicted pixel values of current sub-block are by same complete point
Boundary pixel in resolution cube in encoded adjacent sub-blocks generates, and predicted value and the residual error of true value will be by conducts later
The input of subsequent entropy coding carries out next step coded treatment.If current pixel value is f (u, v, z), wherein (u, v) and z distinguish table
Show the coordinate of this in entire sequence of grey level spatially and temporally, by the reconstructed value in encoded adjacent blockInto
Row prediction:
Wherein aK, lFor predictive coefficient, k, l are the coordinate of reference pixel.The true value of current pixel and the error of predicted value
Are as follows:
Optimization aim is considered when selecting predict the reference block in order to reach compression effectiveness most preferably for each encoding block:
min{R+λ·D}
R is indicated needed for being encoded all relevant informations (as referred to block message, prediction residual etc.) using current predictive method
Bit number, D are using the distortion after current predictive method coding, and λ is Lagrange multiplier, for adjusting between code rate and distortion
Relationship.It should be noted that the prediction residual of current block and reference block is not necessarily minimum, but the cost letter after entropy coding
Numerical value is centainly minimum.
Second of prediction mode is block prediction in the time domain, uses the encoded sub-block in neighbouring full resolution cube
Predict the pixel value in present encoding sub-block.Since the motion information of pulse signal reflection has stronger correlation in the time domain,
Therefore Efficient Compression may be implemented by the prediction in time domain.
Concrete implementation method is that one best is found in encoded both full-pixel cube before for current sub-block
With sub-block, the displacement of best match sub-block to present encoding sub-block is motion vector, and the difference between them is prediction residual.Together
Sample, in the prediction of time domain block, the present invention is also to minimize using encoding motion vector MV when current matching block and prediction
Bit number required for the information such as residual error is optimization aim.
S4, transform and quantization
The distribution of pulse signal spatially is more mixed and disorderly, can make to switch to becoming in the dispersed distribution of spatial domain by transformation
The Relatively centralized distribution for changing domain, scans in conjunction with quantization and " z ", can further promote compression efficiency.
Conversion section can be available there are many specific mapping mode, such as discrete cosine transform (DCT, Discrete
Cosine Transform), discrete sine transform (DST, Discrete Sine Transform) etc..With it is most-often used from
For dissipating cosine transform, the u row xth column element of two-dimensional dct transform Matrix C be may be expressed as:
Wherein
The coefficient matrix of the two-dimensional dct of the signal matrix f of N × N can be expressed as with matrix multiplication
In quantizing process, according to actual scene, compression ratio can be controlled by adjusting quantization step, quantization step is got over
Greatly, compression ratio is bigger, but also brings along bigger error.Quantizing process can state are as follows:
Wherein,It is the element of u row v column in the matrix of DCT unit output,It is its corresponding quantized value.
S5, entropy coding
The last present invention is obtained using before being encoded based on adaptive binary arithmetic coding (CABAC) method of context
The various information arrived.Relate generally to dualization, three parts of context modeling and arithmetic coding.
The optional method of dualization has truncation this dualization (TR) of Lay, K rank Exp-Golomb dualization (EGK) and fixed length two
Memberization (FL) etc. can adjust in practical application the characteristics of as needed or different sequences.By taking fixed length binarization as an example, it is assumed that certain
The value of one given syntactic element is x, and 0≤x≤Max, then is directly converted to binary number rule using decimal number to obtain x
Fixed length two-value symbol string, length
In context model, encoded symbolic information in the adjacent block of present encoding block can be used as present encoding
The context of symbol in block.Adaptive updates are carried out to the variable of probabilistic model after each binary character coding.
It is finally binary arithmetic coding, based on the mode of recurrence interval division, saves coding section in a recursive process
And interval limit.Using adaptive probabilistic model, to each binary code stream root after current syntax element binarization
Arithmetic coding is carried out according to its probabilistic model parameter.The output of arithmetic coding is final code stream.
Embodiment 2
As shown in figure 5, being the structural schematic diagram of lossy compression coded system 20 provided by the invention, comprising:
Conversion module 21, for original burst signal to be converted to sequence of grey level；
Block division module 22 obtains several sub-blocks for carrying out block division to the sequence of grey level；
Prediction module 23 obtains predicted pixel values for being predicted each sub-block, calculates true value and prediction
The residual error of pixel value；
Transform and quantization module 24 carries out transform and quantization operation for the information to the residual error.
Entropy code module 25, information, predictive information and the residual information for being divided described piece by entropy coding are into one
Step compression, and it is stored as binary code stream.
The foregoing is only a preferred embodiment of the present invention, but scope of protection of the present invention is not limited thereto,
In the technical scope disclosed by the present invention, any changes or substitutions that can be easily thought of by anyone skilled in the art,
It should be covered by the protection scope of the present invention.Therefore, protection scope of the present invention should be with the protection model of the claim
Subject to enclosing.
Claims (8)
1. a kind of compressing pulse trains method characterized by comprising
Original burst signal is converted into sequence of grey level；
Block division is carried out to the sequence of grey level, obtains several sub-blocks；
Each sub-block is predicted to obtain predicted pixel values, calculates the residual error of true value and predicted pixel values；
Transform and quantization operation is carried out to the information of the residual error；
Information, predictive information and the residual information that described piece divides further are compressed by entropy coding, and are stored as binary system
Code stream.
2. a kind of compressing pulse trains method according to claim 1, which is characterized in that
It is described that original pulse sequence is converted into sequence of grey level, comprising:
By between the original address added-time, the pulse signal of polar format storage be reduced to the pulse image of 1 bit-depth；
Every continuous several above-mentioned pulse images of frame are synthesized to the single channel image of a frame multiple bit-depth.
3. a kind of compressing pulse trains method according to claim 1, which is characterized in that
Described piece of division include:
The sequence of grey level is divided into the cube of multiple full resolutions, the spatial resolution of each cube in time
It is identical as sequence of grey level；
Each full resolution cube is spatially divided into smaller sub-block.
4. a kind of compressing pulse trains method according to claim 1, which is characterized in that
The block prediction being predicted as in airspace, the predicted pixel values are by encoded adjacent in same full resolution cube
Boundary pixel in sub-block generates.
5. a kind of compressing pulse trains method according to claim 1, which is characterized in that
The block prediction being predicted as in time domain, uses the encoded sub-block prediction present encoding in neighbouring full resolution cube
Pixel value in sub-block.
6. a kind of compressing pulse trains method according to claim 1, which is characterized in that
The entropy coding carries out adaptive binary calculation by predictive information and residual information of the context model to each sub-block
Art coding.
7. a kind of compressing pulse trains method according to claim 6, which is characterized in that
The binary arithmetic coding saves coding section and section based on the mode of recurrence interval division in a recursive process
Lower limit；It is general according to it to each binary file after current syntax element binarization using adaptive probabilistic model
Rate model parameter carries out arithmetic coding.
8. a kind of compressing pulse trains system characterized by comprising
Conversion module, for original burst signal to be converted to sequence of grey level；
Block division module obtains several sub-blocks for carrying out block division to the sequence of grey level；
Prediction module obtains predicted pixel values for being predicted each sub-block, calculates true value and predicted pixel values
Residual error；
Transform and quantization module carries out transform and quantization operation for the information to the residual error.
Entropy code module, for information, predictive information and the residual information that described piece divides further to be compressed by entropy coding,
And it is stored as binary code stream.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811217843.1A CN109474825B (en) | 2018-10-18 | 2018-10-18 | Pulse sequence compression method and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811217843.1A CN109474825B (en) | 2018-10-18 | 2018-10-18 | Pulse sequence compression method and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109474825A true CN109474825A (en) | 2019-03-15 |
| CN109474825B CN109474825B (en) | 2020-07-10 |
Family
ID=65664230
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811217843.1A Active CN109474825B (en) | 2018-10-18 | 2018-10-18 | Pulse sequence compression method and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109474825B (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114155560A (en) * | 2022-02-08 | 2022-03-08 | 成都考拉悠然科技有限公司 | Light weight method of high-resolution human body posture estimation model based on space dimension reduction |
| CN114819121A (en) * | 2022-03-28 | 2022-07-29 | 中国科学院自动化研究所 | Signal processing device and signal processing method based on impulse neural network |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1455600A (en) * | 2003-05-19 | 2003-11-12 | 北京工业大学 | Interframe predicting method based on adjacent pixel prediction |
| US20130016779A1 (en) * | 2004-06-28 | 2013-01-17 | Google Inc. | Video compression and encoding method |
| CN107330843A (en) * | 2017-06-15 | 2017-11-07 | 深圳大学 | A kind of gray level image coding hidden method and device, coding/decoding method and device |
| CN108632630A (en) * | 2018-05-28 | 2018-10-09 | 中国科学技术大学 | A kind of bi-level image coding method of combination bit arithmetic and probabilistic forecasting |
-
2018
- 2018-10-18 CN CN201811217843.1A patent/CN109474825B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1455600A (en) * | 2003-05-19 | 2003-11-12 | 北京工业大学 | Interframe predicting method based on adjacent pixel prediction |
| US20130016779A1 (en) * | 2004-06-28 | 2013-01-17 | Google Inc. | Video compression and encoding method |
| CN107330843A (en) * | 2017-06-15 | 2017-11-07 | 深圳大学 | A kind of gray level image coding hidden method and device, coding/decoding method and device |
| CN108632630A (en) * | 2018-05-28 | 2018-10-09 | 中国科学技术大学 | A kind of bi-level image coding method of combination bit arithmetic and probabilistic forecasting |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114155560A (en) * | 2022-02-08 | 2022-03-08 | 成都考拉悠然科技有限公司 | Light weight method of high-resolution human body posture estimation model based on space dimension reduction |
| CN114819121A (en) * | 2022-03-28 | 2022-07-29 | 中国科学院自动化研究所 | Signal processing device and signal processing method based on impulse neural network |
| CN114819121B (en) * | 2022-03-28 | 2022-09-27 | 中国科学院自动化研究所 | Signal processing device and signal processing method based on impulse neural network |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109474825B (en) | 2020-07-10 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105432077A (en) | Adjusting quantization/scaling and inverse quantization/scaling when switching color spaces | |
| CN105230023A (en) | The self adaptation of color space, color samples rate and/or bit-depth switches | |
| KR101724555B1 (en) | Method and Apparatus for Encoding and Method and Apparatus for Decoding | |
| KR20220145405A (en) | An image encoding/decoding method and apparatus for performing feature quantization/inverse quantization, and a recording medium for storing a bitstream | |
| JP2018502480A (en) | System and method for mask-based processing of blocks of digital images | |
| JP7434604B2 (en) | Content-adaptive online training using image replacement in neural image compression | |
| EP4283993A1 (en) | Video coding and decoding and model training method and apparatus | |
| CN114787870A (en) | Method and apparatus for inter-picture prediction with virtual reference pictures for video coding | |
| JP3794749B2 (en) | Video signal encoding device | |
| CN109474825A (en) | A kind of compressing pulse trains method and system | |
| JP2023532397A (en) | Content-adaptive online training method, apparatus and computer program for post-filtering | |
| CN109379590B (en) | Pulse sequence compression method and system | |
| WO2022266955A1 (en) | Image decoding method and apparatus, image processing method and apparatus, and device | |
| EP3515082B1 (en) | Server device for streaming video content and client device for receiving and rendering video content | |
| US20230281982A1 (en) | Apparatus and method for transceving feature map extracted using mpeg-vcm | |
| Lauga et al. | Segmentation-based optimized tone mapping for high dynamic range image and video coding | |
| KR20010019704A (en) | Macroblock-based object-oriented coding method of image sequence having a stationary background | |
| JP2009055236A (en) | Video coder and method | |
| Fouad | A lossless image compression using integer wavelet transform with a simplified median-edge detector algorithm | |
| JP2024507924A (en) | Hierarchical audio/video or image compression methods and devices | |
| Yin et al. | A co-prediction-based compression scheme for correlated images | |
| US20050129110A1 (en) | Coding and decoding method and device | |
| KR101757464B1 (en) | Method and Apparatus for Encoding and Method and Apparatus for Decoding | |
| KR20000032761A (en) | Method for dividing image into areas in macro block and constructing background mosaic | |
| CN115103188B (en) | SVC error concealment method, model training method, system and equipment |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |