CN111641825B - 3D denoising method and denoising device embedded into HEVC (high efficiency video coding) coding process - Google Patents

3D denoising method and denoising device embedded into HEVC (high efficiency video coding) coding process Download PDF

Info

Publication number
CN111641825B
CN111641825B CN202010522425.4A CN202010522425A CN111641825B CN 111641825 B CN111641825 B CN 111641825B CN 202010522425 A CN202010522425 A CN 202010522425A CN 111641825 B CN111641825 B CN 111641825B
Authority
CN
China
Prior art keywords
coding
residual
denoising
image
time
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
Application number
CN202010522425.4A
Other languages
Chinese (zh)
Other versions
CN111641825A (en
Inventor
章旭东
田景军
居彩霞
陈佳男
齐能
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shanghai Fullhan Microelectronics Co ltd
Original Assignee
Shanghai Fullhan Microelectronics Co ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shanghai Fullhan Microelectronics Co ltd filed Critical Shanghai Fullhan Microelectronics Co ltd
Priority to CN202010522425.4A priority Critical patent/CN111641825B/en
Publication of CN111641825A publication Critical patent/CN111641825A/en
Application granted granted Critical
Publication of CN111641825B publication Critical patent/CN111641825B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Images

Classifications

    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
    • H04N19/102—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or selection affected or controlled by the adaptive coding
    • H04N19/117—Filters, e.g. for pre-processing or post-processing
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
    • H04N19/80—Details of filtering operations specially adapted for video compression, e.g. for pixel interpolation

Landscapes

  • Engineering & Computer Science (AREA)
  • Multimedia (AREA)
  • Signal Processing (AREA)
  • Compression Or Coding Systems Of Tv Signals (AREA)

Abstract

The invention relates to the technical field of image compression, and provides a 3D denoising method embedded in an HEVC (high efficiency video coding) process, which comprises the following steps: respectively calculating a time domain coding residual error and a space domain coding residual error of an input image; weighting and summing the two to obtain a fused time-space domain coding residual error; and setting a threshold value of a filter according to the time-space domain coding residual error so as to realize the denoising of the input image. The invention has the advantages that no additional storage and complex calculation units are added on the hardware realization, and simultaneously, the low-delay requirement of the coding path is ensured.

Description

3D denoising method and denoising device embedded into HEVC (high efficiency video coding) coding process
Technical Field
The invention relates to the field of image processing, in particular to a 3D denoising method and a denoising device embedded in an HEVC (high efficiency video coding) process.
Background
HEVC is an abbreviation for High Efficiency Video Coding, a relatively new Video compression standard. The new generation of HEVC standard was developed with the aim of replacing the h.264/AVC coding standard. At present, a new generation of more efficient video coding HEVC standard is used as a mainstream video standard, mainly for consumer application scenarios, where the data size is particularly large, the image categories are various, and the scene complexity is high, and the requirements for technical indexes are that of high frame rate, high definition, and high image quality. However, for application scenarios like video monitoring, low frame rate such as low light intensity noise, etc., but with requirements for denoising, no corresponding processing means and method are provided.
In the transmission process of video image signals, various noises are caused by environment, a video acquisition process, a circuit system, a transmission channel and the like, so that a denoising processing unit is required. The denoising method also can be different for different noise types and noise intensities. However, the application of denoising aiming at the encoder, especially the application of a video encoder with low code rate, has the particularity.
Image denoising is always the most basic problem in digital image processing, is also a very key technology, and is always a difficult problem in the field of image processing. In the field of image processing, the quality of image denoising directly influences the processing results of the processes of subsequent image edge detection, feature extraction, image segmentation, pattern recognition and the like.
The video signal entering the encoder is generally not the most original signal (Raw source) acquired by the image acquisition equipment, but is a video signal which is preprocessed by a plurality of stages of preprocessing units (3A and the like), wherein the video signal also comprises 2D/3D denoising processing, and the video signal has a certain inhibition effect on various noises. Therefore, the video signal entering the video encoder can present different video styles along with different influences of various levels of preprocessing units, particularly denoising unit methods, denoising strength, sharpening strength and the like, and completely different encoding efficiencies are caused. The hardware encoder designed according to the existing standardized denoising (2D/3D) method cannot adapt to the input video signals of different video styles well.
Disclosure of Invention
The invention aims to provide a 3D denoising method embedded in an HEVC (high efficiency video coding) process, which comprises the following steps: respectively calculating a time domain coding residual error and a space domain coding residual error of an input image; weighting and summing the two to obtain a fused time-space domain coding residual error; and setting a threshold value of a filter according to the time-space domain coding residual error so as to realize the denoising of the input image.
In the 3D denoising method embedded in the HEVC coding process, in the weighted summation process, the weight w is calculated by the following formula:
w＝a×log(SAD/N+offset)+b
here, SAD is the output value of the encoded integer pixel, offset is the quantization coefficient threshold, a is the additive factor, and b is the multiplicative factor.
The 3D denoising method embedded in the HEVC coding process described above, wherein the calculation formula of the time-domain coding residual is:
Figure BDA0002532562070000021
wherein k istpIs the kalman coefficient, d is the prediction residual of the image,
Figure BDA0002532562070000022
is the average of the prediction residuals.
The 3D denoising method embedded in the HEVC coding process described above, wherein the calculation formula of the time-domain coding residual is:
Figure BDA0002532562070000023
wherein k isspD is the predicted residual of the image, SnFor image values that contain additive noise,
Figure BDA0002532562070000024
is SnIs measured.
The 3D denoising method embedded in the HEVC coding process is described above, wherein the kalman coefficient is calculated by the following formula:
Figure BDA0002532562070000025
wherein the content of the first and second substances,
Figure BDA0002532562070000026
to update the factor, δnIs the variance of the additive noise n.
The 3D denoising method embedded in the HEVC coding process described above, wherein the optimal linear fitting coefficient is calculated by the following formula:
Figure BDA0002532562070000027
wherein, deltanIs the variance, δ, of additive noise nSnFor image values S containing additive noisenThe variance of (c).
Another object of the present invention is to provide a 3D denoising module embedded in an HEVC coding process, comprising:
the spatial domain denoising unit is used for receiving input video information and calculating a spatial domain coding residual error according to the video information and the mean value of the video information;
the time domain denoising unit is used for receiving input video information and a predicted value of a video, and calculating a time domain coding residual according to a prediction residual and an average value of the prediction residual;
the fusion and filtering unit determines a weight and performs weighted summation on the space domain coding residual and the time domain coding residual so as to obtain a time-space domain coding residual after the time domain and the space domain are fused; and determining a threshold value of an early filter according to the time-space domain coding residual error, and outputting the denoised image information.
The invention also provides a storage medium on which an executable program is stored, which when running implements the above method
Compared with the prior art, the technical scheme of the invention firstly uses the spatial coding residual and the time domain coding residual which are respectively deduced based on the SSIM objective evaluation index and the Kalman optimal estimation method, then uses SAD calculation cost as a weight judgment basis, weights and sums the spatial coding residual and the time domain coding residual to obtain the time-spatial coding residual which synthesizes time domain and spatial domain characteristics, determines the threshold value of a denoising filter according to the time-spatial coding residual, fully synthesizes the structural characteristics of an encoder, does not need to add extra storage and complex calculation units during hardware realization, and simultaneously ensures the low-delay requirement of a coding channel.
Drawings
Those skilled in the art will appreciate that the following drawings merely illustrate some embodiments of the invention and that other embodiments (drawings) of the same nature can be obtained by those skilled in the art without the exercise of inventive faculty.
Fig. 1 is a schematic diagram of the position of a denoising module according to the present invention in the HEVC encoder architecture;
FIG. 2 is a block diagram of an embodiment of a denoising module according to the present invention.
Detailed Description
In order to make the objects and features of the present invention more comprehensible, embodiments of the present invention are described in detail below with reference to the accompanying drawings. This invention may, however, be embodied in different forms and should not be construed as limited to the embodiments set forth herein. Also, the embodiments and features of the embodiments in the present application are allowed to be combined with or substituted for each other without conflict. The advantages and features of the present invention will become more apparent in conjunction with the following description.
It is to be noted that the drawings are in a very simplified form and are not to precise scale, which is merely for the purpose of facilitating and distinctly claiming the embodiments of the present invention.
It should also be noted that the numbering of the steps in the present invention is for ease of reference and not for limitation of the order of the steps. Specific language will be used herein to describe the particular sequence of steps which is required.
Image denoising refers to the process of reducing noise in a digital image. In reality, digital images are affected by interference of imaging equipment and external environment noise during digitization and transmission, and images input to an encoder are all images containing noise.
The noise of digital images is mainly classified into three types: additive noise, typically associated with the channel over which the image is transmitted; multiplicative noise, typically associated with the video source (i.e., the imaging device) itself; quantization noise, which is essentially an error resulting from digitizing (quantizing) an image signal. The purpose of denoising is to separate the noise from the noisy image to recover the original image.
As described in the background, the video signal entering the hardware video encoder is not the most primitive video signal sensed by the video capture device, but rather is a signal that incorporates various additive, multiplicative, and quantization noises. In different processing devices or device environments, the characteristics of the mixed noise are different, and particularly, if a pre-denoising unit exists, under the influence of the denoising strength, the sharpening strength and the like of the pre-denoising unit, if a hardware video encoder is designed according to an inherent standard denoising method, completely different encoding efficiencies are caused, and the universality of the encoder is influenced.
The core idea of the invention is that after the video signal enters the denoising module disclosed by the invention, namely before the traditional video coding and quantization steps, the coding residual errors are respectively calculated for the time domain and the space domain, then the coding residual errors and the coding residual errors are weighted and summed to obtain a relatively balanced filtering threshold value between the time domain and the space domain, and then the threshold value is applied to a noise filter so as to improve the quality of a coded image and the compression efficiency of coding. The method makes full use of the hardware structure of the encoder, does not need to add extra storage and complex computing units in implementation, and can meet the requirement of low delay of an encoding path.
Fig. 1 is a schematic diagram illustrating the position of a denoising module according to the present invention in the whole encoding process. In a standard HEVC coding process, predicted values between frames and in frames are used as adjustment parameters; the parameters are fed back to the input end of the video and are integrated with the input video to obtain the next frame of image to be coded; on the basis of this image, it is transformed and quantized, and then encoded. The invention embeds the denoising module into the standard HEVC coding process, and performs denoising processing on the image before coding. The denoising module is arranged in front of the transformation and quantization module and is used for denoising the image to be coded of the next frame, so that the transformed and quantized object is minimally influenced by noise.
The denoising module realizes a 3D denoising method embedded in an HEVC (high efficiency video coding) process, and the method respectively calculates a time domain coding residual and a space domain coding residual of an input image; weighting and summing the two to obtain a fused time-space domain coding residual error; and setting a threshold value of a filter according to the time-space domain coding residual error so as to realize the denoising of the input image.
The classical 3D denoising method generally adopts a combination of spatial domain denoising and time domain denoising. However, the spatial domain denoising method is generally completed by using a low-pass filter, and in order to pursue a better filtering effect, the filtering process is also transformed to the frequency domain. For the time domain denoising method, the encoder itself has all the implementation conditions required for time domain prediction. However, there are only two ways to accomplish efficient filtering: first, adaptive filtering; and secondly, based on some reliable objective evaluation index. The adaptive filtering needs to be designed based on the complexity and variability of the actual application scene, and meanwhile, the application of combining the coding characteristics needs to be considered, and on the premise, the design difficulty and the implementation complexity of the adaptive filter are quite high. The inventor considers that no matter what filtering means is adopted, the final filtering result is reliable only by depending on a certain evaluation system, so that the invention adopts a second method, namely, the filtering process is guided based on a certain reliable objective evaluation index so as to ensure the reliability of the filtering result.
The only node which can simultaneously acquire space domain information and time domain information in the encoding process of the encoder is the position for calculating the residual, so the method provided by the patent is to perform denoising processing on the residual, and the calculation of the residual is derived from predicted values, wherein the predicted values comprise intra-frame prediction of a space domain and inter-frame prediction of a time domain.
Specifically, in the 3D denoising method embedded in the HEVC coding process, the spatial coding residual is calculated based on Structural Similarity (SSIM), and the temporal coding residual is calculated based on kalman estimation.
In the conventional video coding standard, a Peak Signal to Noise Ratio (PNSR) is always used as an objective index for evaluating a coded image, and various evaluation indexes of the whole coding system including an integer pixel search cost, a fractional pixel search cost, a mode decision and the like are also constructed based on the PNSR system.
However, the SSIM presenter demonstrates the defects and deficiencies of the PSNR index in the practical application scenario, especially under the influence of noise, and therefore, proposes an index evaluation method based on the SSIM. The invention just applies the index evaluation method based on SSIM to the coding process, and further derives an optimal linear fitting coefficient kspWith this coefficient, the spatial coding residual can be further calculated.
First, for an image S to be encodednFor example, the following relationships are given:
Sn＝S+n
wherein S is the original without noiseAn image, n is an additive noise component in the image and satisfies a distribution
Figure BDA0002532562070000061
White gaussian noise.
Denoising result SdenAnd an image S to be encodednHas the following linear relationship in the spatial domain:
Figure BDA0002532562070000062
wherein k is a fitting coefficient,
Figure BDA0002532562070000063
is SnIs measured.
The fitting coefficient k may be derived from the calculation formula of SSIM.
SSIM is measured based on three comparisons between samples x and y: brightness (luminance), contrast (contrast), and structure (structure). The calculation formula of SSIM can be deduced through the xy relation functions of the three components as follows:
Figure BDA0002532562070000064
wherein, muxIs the mean value of x, μyIs the mean value of y, δ2 xIs the variance of x, δ2 yIs the variance of y, δxyCovariance of x and y, c1 ═ k1L)2，c2＝(k2L)2Is two constants, L is the range of pixel values, k1＝0.01，k20.03 is a common default value. Combining with the formula (3), x in the formula (1) is SnY is Sden。
In view of the uncorrelated between the noise n and the input video Sn, equation (1) can be approximated as:
Figure BDA0002532562070000065
here, in order to obtain an optimal SSIM value (SSIM ═ 1.0), an optimal linear fitting coefficient k in the air space is obtainedspThe value:
Figure BDA0002532562070000071
wherein the content of the first and second substances,
Figure BDA0002532562070000072
kalman filtering is an algorithm for seeking a set of recursive estimation by taking minimum mean square error as an optimal criterion for estimation, and the basic idea is as follows: and updating the estimation of the state variable by using the estimation value of the previous moment and the observation value of the current moment by using a state space model of the signal and the noise, and obtaining the estimation value of the current moment. It is suitable for real-time processing and computer operation.
From the time domain search results, SdenInitial state of
Figure BDA0002532562070000073
The following relationship is satisfied:
Figure BDA0002532562070000074
wherein p is a time domain search prediction image, d is an image prediction residual value, and distribution is assumed to be satisfied
Figure BDA0002532562070000075
From kalman estimation, one can derive:
Figure BDA0002532562070000076
then, the Kalman coefficient
Figure BDA0002532562070000077
Wherein the content of the first and second substances,
Figure BDA0002532562070000078
a Kalman updating process:
Figure BDA0002532562070000079
using the Kalman coefficient ktpThe time-domain coded residual may be further calculated.
By combining the characteristics of the foreground and the background in the coding scene, the background can fully utilize the advantages of time-domain filtering, smooth noise and reduce the coding rate; and for moving foreground objects, the phenomenon of relatively serious ghost can be generated by excessively depending on time domain processing, and at the moment, the spatial filtering based on SSIM can smooth noise while protecting moving edges and object details on the premise of effectively ensuring subjective quality.
Therefore, after the time-space domain residual calculation and output are respectively completed, the time domain and the space domain need to be fused.
Firstly, the optimal linear fitting coefficient kspThe calculation process incorporated into the coded residual includes:
Figure BDA00025325620700000710
wherein d ═ SnP, p is the predicted value.
Then, the Kalman coefficient k is calculatedtpThe calculation process incorporated into the coded residual includes:
difftp＝Sden-p＝ktp*d+(1-ktp)*d
wherein d ═ SnP, p is the predicted value.
Finally, diff is addedspAnd difftpCarrying out weighted summation to complete the fusion of time domain and space domain denoising:
diff＝w*diffsp+(1-w)*difftp
wherein w is the weight value of the time-space domain.
Specifically, in the present invention, the weight value w uses the output value sad (sum of Absolute difference) of the coded integer pixel search as the basis for determination.
By the formula
Figure BDA0002532562070000081
It can be seen that the SAD index basically reflects the physical meaning of the time domain residual mean part in the motion matching process based on blocks, and can effectively represent the degree of motion of an object.
The invention adopts the following formula to calculate the weight w of the time-space domain:
w＝a*log(SAD/N+offset)+b
here, SAD is the output value of the encoded integer pixel, offset is the quantization coefficient threshold, a is the additive factor, and b is the multiplicative factor.
The method carries out denoising processing based on the residual error, can carry out filtering processing in a space domain and a time domain according to the space domain residual error and the time domain residual error respectively, integrates the advantages of the space domain residual error and the time domain residual error (the space domain filtering has strong smoothing capacity, but the details are easy to lose, the time domain retains the details, but the ghost phenomenon is possibly generated on the edge of a moving object), then fuses the space domain residual error and the time domain residual error, balances the advantages of the smoothing filtering and the detail tracking, and can obtain more ideal images for different input video signals.
Fig. 2 is a block diagram illustrating a 3D denoising module embedded in an HEVC coding process.
And the spatial domain denoising unit is used for receiving the input video information and calculating a spatial domain coding residual error according to the video information and the mean value of the video information.
And the time domain denoising unit is used for receiving the input video information and the predicted value of the video and calculating the time domain coding residual according to the prediction residual and the average value of the prediction residual.
The fusion and filtering unit determines a weight and performs weighted summation on the space domain coding residual and the time domain coding residual so as to obtain a time-space domain coding residual after the time domain and the space domain are fused; and determining a threshold value of an early filter according to the time-space domain coding residual error, and outputting the denoised image information.
The present invention also provides a storage medium, on which an executable program is stored, and when the executable program is called by a CPU and runs, the lossless encoding method or the image compression method or the image decompression method (partial or full method) can be implemented. The storage medium includes various physical storage media such as a floppy disk, a hard disk, a flash disk and the like, and the physical relationship with the CPU is not limited to the same device, that is, the storage medium and the CPU may exchange instructions and data through wired/wireless.
The 3D denoising method and the denoising module embedded into the HEVC coding process fuse the coding residual of the time domain and the space domain to obtain a relatively balanced time-space domain coding residual which is used for guiding noise filtering, no additional storage and complex computing unit are added in hardware implementation, and the low-delay requirement of a coding channel is ensured.
It will be apparent to those skilled in the art that various changes and modifications may be made in the invention without departing from the spirit and scope of the invention. Thus, if such modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalents, the present invention is also intended to include such modifications and variations.

Claims (3)

1. A3D denoising method embedded in an HEVC (high efficiency video coding) process is characterized by comprising the following steps: respectively calculating a time domain coding residual error and a space domain coding residual error of an input image; weighting and summing the two to obtain a fused time-space domain coding residual error; setting a threshold value of a filter according to the time-space domain coding residual error so as to realize the denoising of the input image;
in the weighted summation process, the weight w is calculated by the following formula:
w＝a×log(SAD/N+offset)+b
wherein, SAD is the output value of the coding integer pixel, N is the pixel number of the coding integer pixel, offset is the quantization coefficient threshold, a is multiplicative factor, and b is additive factor;
the calculation formula of the time domain coding residual is as follows:
Figure FDA0003492199160000011
wherein k istpIs the kalman coefficient, d is the prediction residual of the image,
Figure FDA0003492199160000012
is the average of the prediction residuals;
the calculation formula of the spatial coding residual is as follows:
Figure FDA0003492199160000013
wherein k isspD is the predicted residual of the image, SnFor image values that contain additive noise,
Figure FDA0003492199160000014
is SnThe mean value of (a);
the kalman coefficient is calculated by the following equation:
Figure FDA0003492199160000015
wherein the content of the first and second substances,
Figure FDA0003492199160000016
to update the factor, δnIs the variance of additive noise n;
the optimal linear fit coefficient is calculated by:
Figure FDA0003492199160000017
wherein, deltanTo addVariance of the sexual noise n, δSnFor image values S containing additive noisenThe variance of (c).
2. A3D denoising module embedded in an HEVC coding process, comprising:
the spatial domain denoising unit is used for receiving input video information and calculating a spatial domain coding residual error according to the video information and the mean value of the video information;
the time domain denoising unit is used for receiving input video information and a predicted value of a video, and calculating a time domain coding residual according to a prediction residual and an average value of the prediction residual;
the fusion and filtering unit determines a weight and performs weighted summation on the space domain coding residual and the time domain coding residual so as to obtain a time-space domain coding residual after the time domain and the space domain are fused; determining a threshold value of a filter according to the time-space domain coding residual error, and outputting denoised image information;
in the weighted summation process, the weight w is calculated by the following formula:
w＝a×log(SAD/N+offset)+b
wherein, SAD is the output value of the coding integer pixel, N is the pixel number of the coding integer pixel, offset is the quantization coefficient threshold, a is multiplicative factor, and b is additive factor;
the calculation formula of the time-domain coding residual is as follows:
Figure FDA0003492199160000021
wherein k istpIs the kalman coefficient, d is the prediction residual of the image,
Figure FDA0003492199160000022
is the average of the prediction residuals;
the calculation formula of the spatial coding residual is as follows:
Figure FDA0003492199160000023
wherein k isspD is the predicted residual of the image, SnFor image values that contain additive noise,
Figure FDA0003492199160000024
is SnThe mean value of (a);
the kalman coefficient is calculated by the following equation:
Figure FDA0003492199160000025
wherein the content of the first and second substances,
Figure FDA0003492199160000026
to update the factor, δnIs the variance of additive noise n;
the optimal linear fit coefficient is calculated by:
Figure FDA0003492199160000027
wherein, deltanIs the variance, δ, of additive noise nSnFor image values S containing additive noisenThe variance of (c).
3. A storage medium having stored thereon an executable program which, when executed, implements the method of claim 1.
CN202010522425.4A 2020-06-09 2020-06-09 3D denoising method and denoising device embedded into HEVC (high efficiency video coding) coding process Active CN111641825B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN202010522425.4A CN111641825B (en) 2020-06-09 2020-06-09 3D denoising method and denoising device embedded into HEVC (high efficiency video coding) coding process

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN202010522425.4A CN111641825B (en) 2020-06-09 2020-06-09 3D denoising method and denoising device embedded into HEVC (high efficiency video coding) coding process

Publications (2)

Publication Number Publication Date
CN111641825A CN111641825A (en) 2020-09-08
CN111641825B true CN111641825B (en) 2022-04-05

Family

ID=72333377

Family Applications (1)

Application Number Title Priority Date Filing Date
CN202010522425.4A Active CN111641825B (en) 2020-06-09 2020-06-09 3D denoising method and denoising device embedded into HEVC (high efficiency video coding) coding process

Country Status (1)

Country Link
CN (1) CN111641825B (en)

Families Citing this family (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN114022705B (en) * 2021-10-29 2023-08-04 电子科技大学 Self-adaptive target detection method based on scene complexity pre-classification
CN114567782B (en) * 2022-04-27 2022-07-12 江苏游隼微电子有限公司 Raw image compression method and device suitable for 3DNR image noise reduction

Family Cites Families (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US9641729B2 (en) * 2012-04-26 2017-05-02 Futurewei Technologies, Inc. System and method for encoder-integrated media denoising
CN103873743B (en) * 2014-03-24 2015-02-25 中国人民解放军国防科学技术大学 Video de-noising method based on structure tensor and Kalman filtering
CN107046648B (en) * 2016-02-05 2019-12-10 芯原微电子(上海)股份有限公司 Device and method for rapidly realizing video noise reduction of embedded HEVC (high efficiency video coding) coding unit
CN106251318B (en) * 2016-09-29 2023-05-23 杭州雄迈集成电路技术股份有限公司 Denoising device and method for sequence image
CN109660813B (en) * 2017-10-12 2021-04-16 上海富瀚微电子股份有限公司 Quantization noise linear fitting method, coding device and coding method

Also Published As

Publication number Publication date
CN111641825A (en) 2020-09-08

Similar Documents

Publication Publication Date Title
CN101390401B (en) Enhanced image/video quality through artifact evaluation
RU2461977C2 (en) Compression and decompression of images
JP6698077B2 (en) Perceptual optimization for model-based video coding
TWI523501B (en) A method and an apparatus for encoding or decoding a video signal
Zhang et al. Low-rank-based nonlocal adaptive loop filter for high-efficiency video compression
KR100578682B1 (en) Hierarchical motion estimation apparatus and method using variable block sizes
KR101528895B1 (en) Method and apparatus for adaptive feature of interest color model parameters estimation
KR100578433B1 (en) Fading estimation/compensation
CN101257630B (en) Video frequency coding method and device combining with three-dimensional filtering
US8369417B2 (en) Optimal denoising for video coding
JP2013513998A (en) Object recognition video coding strategy
WO2018117896A1 (en) Low complexity mixed domain collaborative in-loop filter for lossy video coding
CN111641825B (en) 3D denoising method and denoising device embedded into HEVC (high efficiency video coding) coding process
JPH08307874A (en) Video signal encoding device
WO2002096118A2 (en) Decoding compressed image data
CN115345785A (en) Dim light video enhancement method and system based on multi-scale space-time feature fusion
WO2017004889A1 (en) Jnd factor-based super-pixel gaussian filter pre-processing method
Gullu Weighted constrained one-bit transform based fast block motion estimation
CN116916036A (en) Video compression method, device and system
JP4197695B2 (en) Video encoding method, apparatus, and program
Naman et al. Inter-frame prediction using motion hints
CN117750020A (en) Method, system, equipment and storage medium for learning video coding
Dai et al. SSIM-based rate-distortion optimization in H. 264
US7333543B2 (en) Motion vector estimation method and apparatus thereof
CN115665413A (en) Method for estimating optimal quantization parameter of image compression

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