CN109525847B - Just noticeable distortion model threshold calculation method - Google Patents
Just noticeable distortion model threshold calculation method Download PDFInfo
- Publication number
- CN109525847B CN109525847B CN201811345416.1A CN201811345416A CN109525847B CN 109525847 B CN109525847 B CN 109525847B CN 201811345416 A CN201811345416 A CN 201811345416A CN 109525847 B CN109525847 B CN 109525847B
- Authority
- CN
- China
- Prior art keywords
- dct
- block
- blocks
- texture
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
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
- H04N19/60—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding
- H04N19/625—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using transform coding using discrete cosine transform [DCT]
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
- H04N19/136—Incoming video signal characteristics or properties
- H04N19/137—Motion inside a coding unit, e.g. average field, frame or block difference
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding
- H04N19/134—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals using adaptive coding characterised by the element, parameter or criterion affecting or controlling the adaptive coding
- H04N19/136—Incoming video signal characteristics or properties
- H04N19/14—Coding unit complexity, e.g. amount of activity or edge presence estimation
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N19/00—Methods or arrangements for coding, decoding, compressing or decompressing digital video signals
- H04N19/10—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
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Signal Processing (AREA)
- Physics & Mathematics (AREA)
- Discrete Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
- Image Processing (AREA)
Abstract
The invention relates to a just noticeable distortion model threshold calculation method, which comprises the following steps: performing DCT transformation on the original image, and calculating corresponding brightness self-adaptive module values and spatial contrast sensitivity function module values; the texture blocks of the image are classified more finely by using the frequency energy distribution characteristics of 8 multiplied by 8DCT blocks, a contrast masking factor is obtained, and a contrast masking module value is calculated; extracting texture features of a current image block by using spatial frequency distribution of DCT coefficients, and calculating texture difference between two different blocks to obtain visual perception adjustment factors of the different blocks; and integrating the modules to obtain a final JND threshold value. According to the algorithm provided by the invention, the JND model can accommodate more noises on the premise of ensuring the visual quality. The model can be widely used for perception image/video coding, watermarking, quality evaluation and the like.
Description
Technical Field
The invention relates to the field of image/video coding, in particular to a perceptual visual redundancy estimation model.
Background
With the rapid development of multimedia technology, images/videos gradually become the main carrier for people to acquire information, but people also have to face the challenges of transmission, storage and the like caused by massive image/video data. Therefore, how to efficiently perform image/video coding becomes a research hotspot in the current academic and industrial fields. The traditional image/video coding technology mainly achieves the purpose of improving the coding efficiency by removing spatial domain and temporal domain redundancies, but the visual perception of human eyes is often ignored. In order to further effectively improve the coding efficiency, researchers are dedicated to research on estimation of visual redundancy of images/videos, and a Just Noticeable Distortion (JND) model is a perceptual model capable of effectively representing the visual redundancy.
The types of JND models can be roughly classified into two types: a pixel domain JND model and a DCT domain JND model. This patent focuses on the study of the DCT domain JND model. Because the conventional DCT domain JND model contrast masking only considers the visual masking effect brought by three block types, namely a smooth block, an edge block and a texture block, the influence generated by the texture blocks with different complexity is ignored, and meanwhile, the relation between the blocks of the image is also ignored. Therefore, the accuracy of the conventional DCT domain JND model for visual redundancy estimation needs to be further improved.
Disclosure of Invention
The main purpose of the present invention is to overcome the above-mentioned drawbacks in the prior art, and to provide a method for calculating a just noticeable distortion model threshold, further considering different visual masking effects that may be caused by texture blocks of different complexity, and considering the relation between image blocks and blocks, and providing a more accurate visual redundancy estimation model.
The invention adopts the following technical scheme:
a just noticeable distortion model threshold calculation method is characterized by comprising the following steps:
1) performing DCT transformation on an input image;
2) operating DCT coefficient to calculate brightness self-adaptive module value FLAAnd a spatial contrast sensitivity function module value;
3) further dividing the types of texture blocks by using the frequency energy distribution characteristics of 8 multiplied by 8DCT blocks, and calculating a contrast masking module value;
4) extracting texture features of the current block based on different frequency division of 8 multiplied by 8DCT blocks, calculating texture difference values of the current block and other blocks, and solving visual perception adjustment factors of the blocks;
5) and integrating the brightness self-adaptive module value, the spatial contrast sensitivity function module value, the contrast masking module value and the visual perception adjustment factors of different blocks to obtain the JND threshold of the image.
The brightness adaptive module value and the spatial contrast sensitivity function module value are respectively obtained by the following formulas:
wherein, (i, j) represents the coefficient position of 8 multiplied by 8DCT block, the value range of i and j is integer 0-7, FLARepresenting the value of the luminance adaptation block, FCSF(i, j) represents the spatial contrast sensitivity function module value,represents the average intensity, φ, of an 8 × 8DCT blockiAnd phijNormalization factor, omega, for DCT transformationijThe spatial frequency is represented by a representation of,the angle of direction of the corresponding DCT component, a, b, c, r, s, is constant.
The type of the texture block is further divided according to the frequency distribution of the 8 × 8DCT block coefficients, and the contrast masking module is obtained by the following formula:
where ψ is the contrast masking factor, C (n)1,n2I, j) are the DCT coefficients of the corresponding position, (n)1,n2) ε is set to 0.36 for the position index of the current DCT block in the image.
The contrast masking factor is obtained by the following formula:
wherein texture block energy Et=E+H1E and H1The sum of the absolute values of the DCT coefficients in the MF region and the HF region, respectively.
The visual perception adjustment factor is obtained by the following formula:
αk=1-Sk
wherein SkThe value range of k is an integer for the weight of the kth DCT block in the imageWhere W and H are the width and height of the input image respectively,indicating rounding up.
The weight of the DCT blocks in the image is obtained by the following formula:
where σ is a Gaussian model parameter, dklIs the Euclidean distance, D, of the kth DCT block from the l DCT blockklThe texture difference between the kth DCT block and the l DCT block.
Texture difference DklObtained by the following formula:
Dkl=max(h(Tk,Tl),h(Tl,Tk))
where T denotes the texture feature of the DCT block, h (T)k,Tl) The Hausdorff distance of the texture features of the kth DCT block and the l DCT block.
The Hausdorff distance and DCT block texture features of two different DCT block texture features are obtained by the following formulas:
T={tLF,tMF,tHF}
wherein t isLF,tMFAnd tHFThe sum of the coefficients of the LF, MF and HF regions, respectively, | | · | |, represents a 2 norm.
The JND threshold of the DCT domain is obtained by the following formula:
the JND threshold is the visual perception adjustment factor × spatial contrast sensitivity function module value × luminance adaptive module value × contrast masking module value.
As can be seen from the above description of the present invention, compared with the prior art, the present invention has the following advantages:
1. the method of the invention classifies texture blocks more finely according to the frequency energy of DCT blocks, so that the estimation of contrast masking is more in line with the visual characteristics of human eyes;
2. the method of the invention considers the relation between the image blocks, extracts the texture characteristics of the DCT blocks, obtains the visual perception adjustment factor of the image blocks by calculating the texture difference value of the current block and other blocks, and is beneficial to the accurate estimation of the visual redundancy.
Drawings
FIG. 1 is a main flow chart of the process of the present invention
Fig. 2 is a frequency distribution diagram of 8 × 8DCT blocks.
Detailed Description
The invention is further described below by means of specific embodiments.
The invention provides a method for accurately estimating visual redundancy and calculating a just noticeable distortion model threshold, which comprises the following specific implementation steps as shown in fig. 1:
1) the DCT transform is performed on the input image.
2) And operating the DCT coefficient to calculate a brightness self-adaptive module value and a spatial contrast sensitivity function module value.
3) And further dividing the types of the texture blocks by using the frequency energy distribution characteristics of the 8 multiplied by 8DCT blocks, and calculating a contrast masking module value.
4) Based on different frequency division of 8 multiplied by 8DCT blocks, extracting texture features of the current block, calculating texture difference values of the current block and other blocks, and solving visual perception adjustment factors of the blocks.
5) And integrating the brightness self-adaption module, the spatial contrast sensitivity function module, the contrast masking module and the weight of different blocks to obtain the JND threshold of the image.
Step 1), performing DCT transformation on the input image.
Step 2), calculating a brightness self-adaptive module value FLAAnd spatial contrast sensitivity function module value FCSF(i, j). The method comprises the following specific steps:
wherein (i, j) represents the position of 8 multiplied by 8DCT block coefficient, the value range of i and j is integer 0-7,represents the average intensity, φ, of an 8 × 8DCT blockiAnd phijFor DCT transformationA normalization factor, omegaijThe spatial frequency is represented by a representation of,the a, b, c, r, s are constants for the directional angles of the corresponding DCT components, and have values of 1.33,0.11,0.18, 0.6, and 0.25, respectively.
And 3) calculating a contrast masking module value.
Specifically, the contrast masking factor ψ is calculated as follows:
wherein texture block energy Et=E+H1E and H1The sum of the absolute values of the DCT coefficients in the MF region and the HF region, respectively.
The contrast masking block value F is obtained by the following formulaCM。
Where ψ is the contrast masking factor, C (n)1,n2I, j) are the DCT coefficients of the corresponding position, (n)1,n2) Is the position index of the current DCT block in the image and (i, j) is the coordinate position of the DCT coefficient in the current DCT block, epsilon is set to 0.36.
And 4) calculating a perception adjustment factor of each image block.
Specifically, the Hausdorff distance between the texture feature T of the 8 × 8DCT block and the texture features of two different DCT blocks is extracted by the following formula:
T={tLF,tMF,tHF}
wherein t isLF,tMFAnd tHFThe sum of the coefficients of the LF, MF and HF regions, respectively,i | · | represents a 2 norm, and the value ranges of k and l are integersWhere W and H are the width and height of the input image respectively,indicating rounding up.
Obtaining the texture difference value D by the following formulakl:
Dkl=max(h(Tk,Tl),h(Tk,Tl))
Wherein h (T)k,Tl) The Hausdorff distance of the texture features of the kth DCT block and the l DCT block.
The weight of the DCT block in the image is obtained by the following formula:
wherein SkThe weight of the kth DCT block in the image is taken as the sigma is a Gaussian model parameter and is set as 5, dklIs the Euclidean distance, D, of the kth DCT block from the l DCT blockklThe texture difference between the kth DCT block and the l DCT block.
Based on the weight S occupied by DCT blocks in the imagekCalculating a visual perception adjustment factor of the current block, specifically as follows:
αk=1-Sk
wherein SkIs the weight of the kth DCT block in the image. k is an integerWhere W and H are the width and height of the input image respectively,indicating rounding up.
Step 5) obtaining the JND threshold value through the following formula:
the JND threshold is the visual perception adjustment factor × spatial contrast sensitivity function module value × luminance adaptive module value × contrast masking module value, that is:
JND(n1,n2,i,j)=α(n1,n2)·FCSF(n1,n2,i,j)·FLA(n1,n2)·FCM(n1,n2,i,j)。
the above description is only an embodiment of the present invention, but the design concept of the present invention is not limited thereto, and any insubstantial modifications made by using the design concept should fall within the scope of infringing the present invention.
Claims (3)
1. A just noticeable distortion model threshold calculation method is characterized by comprising the following steps:
1) performing DCT transformation on an input image;
2) operating DCT coefficient to calculate brightness self-adaptive module value FLAAnd a spatial contrast sensitivity function module value;
3) the method comprises the following steps of further dividing the types of texture blocks by utilizing the frequency energy distribution characteristics of 8 multiplied by 8DCT blocks, calculating a contrast masking factor and further calculating a contrast masking module value, wherein the contrast masking factor is obtained by the following formula:
wherein texture block energy Et=E+H1E and H1Respectively representing the sum of DCT coefficient absolute values of an MF area and an HF area, (i, j) represents the coefficient position of an 8 multiplied by 8DCT block, and the value range of i and j is an integer of 0-7;
4) extracting texture features of the current block based on different frequency division of 8 multiplied by 8DCT blocks, calculating texture difference values of the current block and other blocks, and solving visual perception adjustment factors of the blocks;
the visual perception adjustment factor is obtained by the following formula:
αk=1-Sk
wherein SkThe value range of k is an integer for the weight of the kth DCT block in the imageWhere W and H are the width and height of the input image respectively,represents rounding up;
the weight of the DCT blocks in the image is obtained by the following formula:
where σ is a Gaussian model parameter, dklIs the Euclidean distance, D, of the kth DCT block from the l DCT blockklThe texture difference value of the kth DCT block and the l DCT block is obtained;
the texture difference value DklObtained by the following formula:
Dkl=max(h(Tk,Tl),h(Tl,Tk))
where T denotes the texture feature of the DCT block, h (T)k,Tl) The Hausdorff distance of texture features of the kth DCT block and the l DCT block is obtained;
the Hausdorff distance and DCT block texture features of two different DCT block texture features are obtained by the following formulas:
T={tLF,tMF,tHF}
wherein t isLF,tMFAnd tHFIs the sum of the coefficients of the LF, MF and HF regions, respectively, | | · | | | represents a 2 norm;
5) integrating a brightness self-adaptive module value, a spatial contrast sensitivity function module value, a contrast masking module value and visual perception adjustment factors of different blocks to obtain a JND threshold value of the image;
wherein the JND threshold of the DCT domain is obtained by the following formula:
the JND threshold is the visual perception adjustment factor × spatial contrast sensitivity function module value × luminance adaptive module value × contrast masking module value.
2. The method of claim 1, wherein the luminance adaptive module value and the spatial contrast sensitivity function module value are obtained by the following formulas:
wherein, FLARepresenting the value of the luminance adaptation block, FCSF(i, j) represents the spatial contrast sensitivity function module value,represents the average intensity, φ, of an 8 × 8DCT blockiAnd phijNormalization factor, omega, for DCT transformationijThe spatial frequency is represented by a representation of,the angle of direction of the corresponding DCT component, a, b, c, r, s, is constant.
3. The method of claim 2, wherein the type of texture block is further divided according to the frequency distribution of 8 x 8DCT block coefficients, and the contrast masking module is obtained by the following formula:
where ψ is the contrast masking factor, C (n)1,n2I, j) are the DCT coefficients of the corresponding position, (n)1,n2) ε is set to 0.36 for the position index of the current DCT block in the image.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811345416.1A CN109525847B (en) | 2018-11-13 | 2018-11-13 | Just noticeable distortion model threshold calculation method |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811345416.1A CN109525847B (en) | 2018-11-13 | 2018-11-13 | Just noticeable distortion model threshold calculation method |
Publications (2)
Publication Number | Publication Date |
---|---|
CN109525847A CN109525847A (en) | 2019-03-26 |
CN109525847B true CN109525847B (en) | 2021-04-30 |
Family
ID=65776416
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811345416.1A Active CN109525847B (en) | 2018-11-13 | 2018-11-13 | Just noticeable distortion model threshold calculation method |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109525847B (en) |
Families Citing this family (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109982022A (en) * | 2019-04-17 | 2019-07-05 | 南京大学 | The video refreshing method of minimum color difference can be examined based on human eye |
CN110062234B (en) * | 2019-04-29 | 2023-03-28 | 同济大学 | Perceptual video coding method based on just noticeable distortion of region |
CN112634278B (en) * | 2020-10-30 | 2022-06-14 | 上海大学 | Super-pixel-based just noticeable distortion method |
CN112437302B (en) * | 2020-11-12 | 2022-09-13 | 深圳大学 | JND prediction method and device for screen content image, computer device and storage medium |
CN112584153B (en) * | 2020-12-15 | 2022-07-01 | 深圳大学 | Video compression method and device based on just noticeable distortion model |
CN112866820B (en) * | 2020-12-31 | 2022-03-08 | 宁波大学科学技术学院 | Robust HDR video watermark embedding and extracting method and system based on JND model and T-QR and storage medium |
CN112967229B (en) * | 2021-02-03 | 2024-04-26 | 杭州电子科技大学 | Method for calculating just-perceived distortion threshold based on video perception characteristic parameter measurement |
CN113192083B (en) * | 2021-05-07 | 2023-06-06 | 宁波大学 | Image just noticeable distortion threshold estimation method based on texture masking effect |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101621708A (en) * | 2009-07-29 | 2010-01-06 | 武汉大学 | Method for computing perceptible distortion of color image based on DCT field |
CN102420988A (en) * | 2011-12-02 | 2012-04-18 | 上海大学 | Multi-view video coding system utilizing visual characteristics |
CN103475881A (en) * | 2013-09-12 | 2013-12-25 | 同济大学 | Image JND threshold value computing method in DCT domain and based on visual attention mechanism |
CN104469386A (en) * | 2014-12-15 | 2015-03-25 | 西安电子科技大学 | Stereoscopic video perception and coding method for just-noticeable error model based on DOF |
Family Cites Families (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US8559511B2 (en) * | 2010-03-30 | 2013-10-15 | Hong Kong Applied Science and Technology Research Institute Company Limited | Method and apparatus for video coding by ABT-based just noticeable difference model |
JP5470415B2 (en) * | 2012-03-30 | 2014-04-16 | Eizo株式会社 | Epsilon filter threshold determination method and low-pass filter coefficient determination method |
-
2018
- 2018-11-13 CN CN201811345416.1A patent/CN109525847B/en active Active
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101621708A (en) * | 2009-07-29 | 2010-01-06 | 武汉大学 | Method for computing perceptible distortion of color image based on DCT field |
CN102420988A (en) * | 2011-12-02 | 2012-04-18 | 上海大学 | Multi-view video coding system utilizing visual characteristics |
CN103475881A (en) * | 2013-09-12 | 2013-12-25 | 同济大学 | Image JND threshold value computing method in DCT domain and based on visual attention mechanism |
CN104469386A (en) * | 2014-12-15 | 2015-03-25 | 西安电子科技大学 | Stereoscopic video perception and coding method for just-noticeable error model based on DOF |
Non-Patent Citations (4)
Title |
---|
A DCT-Based Total JND Profile for Spatiotemporal and Foveated Masking Effects;Sung-Ho Bae;《 IEEE Transactions on Circuits and Systems for Video Technology》;20160308;第27卷(第6期);全文 * |
A Novel DCT-Based JND Model for Luminance Adaptation Effect in DCT Frequency;Sung-Ho Bae;《IEEE Signal Processing Letters》;20130708;第20卷(第9期);全文 * |
A novel direction-based JND model for perceptual HEVC intra coding;Zhipeng Zeng;《2017 International Symposium on Intelligent Signal Processing and Communication Systems (ISPACS)》;20171106;全文 * |
基于纹理分解的变换域JND模型及图像编码方法;郑明魁;《通信学报》;20140630;第35卷(第6期);全文 * |
Also Published As
Publication number | Publication date |
---|---|
CN109525847A (en) | 2019-03-26 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109525847B (en) | Just noticeable distortion model threshold calculation method | |
CN110599409B (en) | Convolutional neural network image denoising method based on multi-scale convolutional groups and parallel | |
Sun et al. | SPSIM: A superpixel-based similarity index for full-reference image quality assessment | |
CN106530237B (en) | A kind of image enchancing method | |
CN109872285B (en) | Retinex low-illumination color image enhancement method based on variational constraint | |
Maalouf et al. | CYCLOP: A stereo color image quality assessment metric | |
CN103049892B (en) | Non-local image denoising method based on similar block matrix rank minimization | |
Zheng et al. | No-reference quality assessment for screen content images based on hybrid region features fusion | |
CN115937216B (en) | Magnetic rotor appearance quality detection method for new energy automobile | |
CN115984134A (en) | Intelligent enhancing method for remote sensing mapping image | |
CN108932699A (en) | Three-dimensional matching reconciliation filtering image denoising method based on transform domain | |
CN117541582B (en) | IGBT insulation quality detection method for high-frequency converter | |
CN110728690B (en) | Matting method for performing edge smoothing processing on image by wavelet transformation | |
CN117474817B (en) | Method for content unification of composite continuous images | |
CN104143203A (en) | Image editing and communication method | |
CN108090914B (en) | Color image segmentation method based on statistical modeling and pixel classification | |
CN111079689B (en) | Fingerprint image enhancement method | |
CN109447952B (en) | Semi-reference image quality evaluation method based on Gabor differential box weighting dimension | |
CN109584190B (en) | Contrast image enhancement method based on wavelet transformation low-frequency histogram limitation | |
CN109544593B (en) | General image noise estimation method based on self-similarity measurement | |
CN110223273B (en) | Image restoration evidence obtaining method combining discrete cosine transform and neural network | |
CN108550119B (en) | Image denoising method combined with edge information | |
CN105894478B (en) | Image de-noising method based on the local order feature of statistics | |
CN113421181B (en) | Information hiding method based on estimated robustness | |
CN111724325B (en) | Trilateral filtering image processing method and trilateral filtering image processing device |
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 |