CN106815877A - A kind of improvement fractal image Underwater Image compression algorithm based on dictionary - Google Patents
A kind of improvement fractal image Underwater Image compression algorithm based on dictionary Download PDFInfo
- Publication number
- CN106815877A CN106815877A CN201710030003.3A CN201710030003A CN106815877A CN 106815877 A CN106815877 A CN 106815877A CN 201710030003 A CN201710030003 A CN 201710030003A CN 106815877 A CN106815877 A CN 106815877A
- Authority
- CN
- China
- Prior art keywords
- dictionary
- block
- image
- fractal
- blocks
- 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.)
- Withdrawn
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T9/00—Image coding
Landscapes
- Engineering & Computer Science (AREA)
- Multimedia (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Compression Of Band Width Or Redundancy In Fax (AREA)
- Compression Or Coding Systems Of Tv Signals (AREA)
Abstract
The present invention relates to a kind of improvement fractal image image compression algorithm based on dictionary, belong to compression of images and process field.Long in view of traditional fractal image mode scramble time, each image carries out coding and dictionary will be regenerated on the basis of original image（Also known as code book）Different images has different dictionaries, even if same piece image, different partitioning schemes also produce different dictionaries, the present invention carries out fractal image using the method for fixed dictionary regarding to the issue above, and generation from dictionary, expand and classification three in terms of improve fractal coding algorithm.The algorithm that the invention is proposed has faster coding rate, and decoding only needs an iteration with the decoding speed being exceedingly fast.In terms of compression ratio, more conventional algorithm is significantly increased.
Description
Technical field
The present invention relates to a kind of improvement fractal image Underwater Image compression algorithm based on dictionary, belong to compression of images with place
Reason field.Long in view of traditional fractal image mode scramble time, each image carries out coding will be on the basis of original image
Regenerate dictionary（Also known as code book）, different images has different dictionaries, even if same piece image, different partitioning schemes
Also different dictionaries are produced, the present invention carries out fractal image using the method for fixed dictionary regarding to the issue above, and from dictionary
Fractal coding algorithm is improved in terms of generation, expansion and classification three.
Background technology
Fractal image is in Fractal Geometry Theory and iterated function system（IFS）On the basis of a kind of image pressure for growing up
Compression method.It takes full advantage of the local self-similarity of generally existing in image, with one group of compression for obtaining in an encoding process
Transformation parameter represents original image, and the attractor and original image for making compressed transform be substantial access to.This conversion is reflected in image
Similitude between part and part.Fractal image have compression ratio it is high, decoding image it is unrelated with resolution ratio and decoding it is fireballing
Advantage, but also have its weak point：It is computationally intensive, coding rate is slow, decoded image quality is not good etc..
Current domestic and foreign scholars are studied the coding rate for how improving fractal image mainly following direction：①
Classification and matching 2. Optimizing Search scope 3. using fixed codebook these three means.Most time-consuming link is during fractal image
Best match block search is carried out in the range of whole domain of definition to each range block, therefore, the scramble time is reduced, it is just necessary
Reduce hunting zone, and ensure in the scope that best matching blocks are located at after reducing, thus most researcher concern it is important that how
Build efficient and rational range block package space.
The method that the present invention is considered as fixing dictionary carries out fractal image.Using dictionary method encode the ring of most critical
Section is the generation of dictionary, expands and classify.The dictionary of conventional method generation not enough enriches, and there is too little or too much asking of classifying
Topic, the size of dictionary only has one kind in addition, may be only available for 4 × 4 segmentations.Therefore, the present invention to the generation of dictionary, expand and point
Class method is improved, and generates three kinds of various sizes of dictionaries, while increasing the rich of dictionary and Underwater Image being fitted
Ying Xing.
The content of the invention
The present invention is divided these images using a large amount of shapes of rich generation of the Fractal Set image different with texture
The various sizes of image block of generation is cut, image block is expanded afterwards, various sizes of image block set of classifying is used as dictionary.
During coding, original image need to only carry out range block segmentation, and to each R after every segmentation, it is special that rule when classifying by dictionary calculates it
Value indicative, then carries out matching search in a certain class of dictionary corresponding to this feature value, and record searches best matching blocks in word
Used as fractal code, the generation of dictionary and coding flow chart are shown in that specification is attached for position, contrast factor and grayscale shift amount in allusion quotation
Shown in Fig. 1.During decoding, each best matching blocks of R blocks in dictionary is found according to fractal code, it is only necessary to linear transformation
Recover.
Realization of the invention comprises the following steps：
Step one：Dictionary creation.
In order that image block miscellaneous can find preferable match block in dictionary, the block in dictionary must be enough
It is abundant and with representativeness widely.The key issue of dictionary creation is how to obtain more rich various image block,
The generation of dictionary of the present invention includes：The generation of Julia Fractal Sets image, dictionary creation and expansion.
Step 2：Coding.
Comprising step in detail below：
（1）It is loaded into the dictionary of generation.
（2）R block threshold values are set, match error threshold 。
（3）Range block segmentation is carried out to image.
（4）Code book when selecting the dictionary of corresponding size as matching search according to range block cut size.
（5）For each R block, its variance is calculated
（6）Step is performed to all R blocks（5）, quantifying fractal code, output forms coding file.
Step 3：Decoding.
Compared with prior art, the beneficial effects of the invention are as follows：Due to being carried out in the affiliated class only in the dictionary of each R block
Search, matching times are greatly decreased than rudimentary algorithm, with faster coding rate, 351 times are averagely improve than rudimentary algorithm.
Additionally, dictionary method decoding only needs an iteration, therefore with the decoding speed being exceedingly fast, 12 times are improve than rudimentary algorithm.In pressure
Aspect is compared in contracting, and the method for the present invention improves 21% than rudimentary algorithm.Therefore, a kind of improvement based on dictionary point shape of the present invention is compiled
Code Underwater Image compression algorithm, can be greatly enhanced encoding and decoding speed, for Underwater Image, can reach at or above basic calculation
The decoding quality of method, for ordinary gamma image, when splitting using 3 × 3 and 4 × 4, dictionary method has performance well, using 6 ×
During 6 segmentation, dictionary method blocking effect is more apparent.
Brief description of the drawings
Fig. 1 dictionary creations and coding flow chart.
Fig. 2 is based on dictionary algorithm coding specific steps.
Fig. 3The numbering of image block each pixel.
Fig. 4The numbering of image block each pixel.
Specific embodiment
The present invention is described in further detail with specific embodiment below in conjunction with the accompanying drawings.
The present invention comprises the following steps：
Step one：Julia collection image is generated.
Julia collection is the set for Complex Plane Fractal point named by French scientist Gaston Julia, by reflecting
Penetrate（、In complex field andIt is constant）By grey iterative generation.Thus mapping can produce a sequence of iterations：, this sequence may be intended to infinite or within a certain range and converge on certain point all the time, make sequence indiffusion
The collection of z values be collectively referred to as Julia collection（Hereinafter referred to as J collection）.On a complex plane, mapping is made：Ethylene absorbent to have
The multiple parameter on boundaryCollection be collectively referred to as Mandelbrot collection（Hereinafter referred to as M collection）.From the definition of J collection and M collection, can be in M collection
Inside take a little as parameterGeneration J collection.
The present invention uses Julia Fractal Sets, realizes generating complicated image with less parameter.Using escape time method
Generation Julia collection images, specifically comprise the steps of：
(1) initialize
Setting imageResolution ratio be, parameter,a, b, p, qIt is the parameter of setting.Escape radius
, maximum iteration.Set up coordinate system：Axle is real number axis,Axle is axis of imaginaries,WithSpan be respectively：, then the point step size in image be：；, orderIn it is every
The pixel value of individual point is 0.
(2) circulate
Selected element,；, complete following iteration：
；；；
；；。
(3) iterative cycles control
Calculate mould：If,, then color is selected, go to and 4. walk, otherwise, go to and 2. walk.
(4) colour
To pointColor, and next pixel is selected, turn the and 2. walk.
(5) coloring is completed a little, obtains J collection images.
To generate various sizes of dictionary, the present invention is respectively provided with image resolution ratio and is、、、, concentrate each point for being taken as parameter c in M by previous step, a large amount of different J collection images can be generated.
Step 2：The generation and expansion of dictionary.
The present invention is no more than to image using sizeSegmentation, therefore, build 3 sizes be respectively, 4 × 4 HesDictionary, be designated as、With.The generation and expansion of dictionary are comprised the following steps that：
（1）Segmentation Julia collection image generates various sizes of dictionary
The linear conversion of intensity value ranges of step one is generated Julia collection images first expands to 0 ~ 255.
1. dictionary is generated
A. will、J collection image segmentations areSub-block；
B. willJ collection image segmentations are、、Sub-block；WillJ collection image segmentations are、、、Sub-block, to more than generation sub-block, use the average method of pixel by its size change over for。
2. dictionary is generated
A. will、J collection image segmentations areSub-block；
B. willJ collection image segmentations are、, 21 × 21 sub-blocks；WillJ collection image segmentations are、12
×12、、, 42 × 42 sub-blocks, to more than generation sub-block, use the average method of pixel by its size change over for。
3. dictionary is generated
A. will、J collection image segmentation is 4 × 4 sub-blocks；
B. willJ collection image segmentations are、、Sub-block；WillJ collection image segmentations are、
、、Sub-block, the sub-block of generation to more than, it is 4 × 4 by its size change over to use the average method of pixel.
（2）Expand dictionary.
For the block for including dictionary is more rich and varied, the present invention expands dictionary in the following ways：
1. prime number is multiplied：Each sub-block in dictionary is multiplied by different prime numbers（After can avoiding being multiplied with certain number from prime number
Block with other number be multiplied after block composition relation）, then to 256 modulus, new image block being generated, finally will newly scheme
As block is added in dictionary.
2. Gaussian noise is added.In view of Underwater Image generally existing noise, and the block after J collection image segmentations is nothing
Noise image block, in order that dictionary is preferably matched with Underwater Image block, during the present invention makes dictionary by adding Gaussian noise
The change of block pixel has more randomness and diversity, to meet Underwater Image coding needs.
3. equilong transformation is carried out.To reduce matching error, with reference to basic fractal image process, the dictionary to generating carries out 8
Equilong transformation is planted, expands dictionary capacity.
（3）Reject the redundancy in dictionary.
1. the constant value block in dictionary is removed（There is a pixel value identical block）.Such piece when fractal image is matched only
Can be with constant value R Block- matchings, and constant value R blocks then can be with any Block- matching（Only need to be by contrast factorIt is set to 0）, therefore it is such
Block need not be stayed in dictionary.
2. the block with linear relationship in dictionary is removed.From fractal image matching process, with linear correlation
The block of relation, matching effect is the same, only exists the difference in match parameter, therefore need to only retain one.
Step 3：The classification of dictionary.
Two factors are considered on the principle of classification of dictionary：
1. classification number number.Less classification is divided for small size dictionary, large scale dictionary is divided into more classification.
2. the number of block is included in each class.The setting of sub-block number meets：After dictionary encoding, compression ratio and peak value are believed
Make an uproar ratio
The two parameters of basic fractal image under same size should be close to or higher than.
Based on above reason, the present invention takes following classification policy respectively to three kinds of various sizes of dictionaries：
（1）Dictionary is classified
1. pressImage block each pixel is numbered, and sees Figure of description 3.
2. formula is pressedCalculate the characteristic value of each block。
（5-1）
In formula,；For numbering is in figurePoint pixel value；It is the pixel average of the block.
3. different classes are set according to characteristic value.Operation rule is 2. walked according to,The characteristic value of image block is not repeated
511 altogether（In the absence of the two-value block of full 0）, each characteristic value correspondenceA class in dictionary.
4. for each block in dictionary, according to its characteristic value, decide whether to put it into its institute by following rules
Corresponding class（Remember that such is）In：
If a.It is empty set, is then directly placed into, otherwise, goes to b.
B. calculate the block withIn already present piece of PSNR（Y-PSNR）If result is respectively less than certain predetermined threshold value
（Set herein）, andMiddle piece of quantity is not up to N, then be put into, and otherwise gives up.
Carry out the screening of this step, the block similarity included in dictionary can be avoided too high, increase in each class block and block it
Between otherness, be conducive to improve encode when matching precision.
5. all pieces in dictionary are carried out with 4. step operation, dictionary classification is completed.
（2）Dictionary is classified
1. will by bilinearity difference approachThe size of image block is reduced into 3 × 3 by 4 × 4 in dictionary.
2. pressDictionary sorting technique is classified.
（3）Dictionary is classified
1. will by bilinearity difference approachIn dictionary the size of image block byIt is reduced into, it is rightImage block each picture
Vegetarian refreshments is numbered, and sees Figure of description 4.
2. it is calculated as follows the characteristic value of each block。
（5-2）
In formula,；；For numbering is in Figure of description 4Point picture
Element value；It is the pixel average of the block.
3. operation rule is 2. walked according to,The characteristic value of image block does not repeat to amount to 8192, each characteristic value
CorrespondenceA class in dictionary.
4. to each block in dictionary according toIn dictionary classification the 4., 5. step operation, complete screening and classify.
Step 4：Coding.
（1）It is loaded into the dictionary according to the generation of step one, two and three.
（2）R block threshold values are set, match error threshold。
（3）Range block segmentation is carried out to image.
（4）Code book when selecting the dictionary of corresponding size as matching search according to range block cut size.
（5）For each R block, its variance is calculated, matching is scanned for as follows, obtain fractal code（Fractal code
The position in dictionary and contrast factor including best matching blocksWith grayscale shift amount）：
If 1., then it is smooth block to regard the R blocks, and any block in optional dictionary is best matching blocks, therefore without search
Fractal code can directly be exported.
If 2.Formula is then pressed according to R block sizesOr
Calculate the characteristic value of the R blocks, the class corresponding to it is selected in dictionary,In scan for matching, complete coding.
（6）Step is performed to all R blocks（5）, quantifying fractal code, output forms coding file.
Step 5：Decoding.
（1）It is loaded into the dictionary by above-mentioned steps generation.
（2）By coding file read position in dictionary of R blocks, each corresponding best matching blocks of R blocks, contrast because
The information such as son, grayscale shift amount.
（3）One and an equal amount of image X of original image is defined, for preserving decoding image.
（4）For each the R block in X, its best matching blocks D, Ran Houyou are found in dictionary according to its fractal codeRecover.
（5）After all R blocks recover, composition decoding image is pieced together.
Claims (9)
1. a kind of improvement fractal image Underwater Image compression algorithm based on dictionary, it is characterised in that：Comprise the following steps：
Step one：Julia collection image is generated；
Step 2：The generation and expansion of dictionary；
Step 3：The classification of dictionary；
Step 4：Coding；
Step 5：Decoding.
2. a kind of improvement fractal image Underwater Image compression algorithm based on dictionary according to claim 1, its feature exists
In：The generation of Julia collection image is comprised the steps of in step one：
(1) initialize
Setting imageResolution ratio be, parameter,a, b, p, qIt is the parameter that can be set, c is multiple parameter.Escape
Ease radius, maximum iteration.Set up coordinate system：Axle is real number axis,Axle is axis of imaginaries,WithValue
Scope is respectively：, then the point step size in image be：；, orderIn each point pixel value be 0；
(2) circulate
Selected element,；, complete following iteration：
；；；
；；；
(3) iterative cycles control
Calculate mould：If,, then color is selected, go to and 4. walk, otherwise, go to and 2. walk；
(4) colour
To pointColor, and next pixel is selected, turn the and 2. walk；
(5) coloring is completed a little, obtains Julia collection images；
To generate various sizes of dictionary, the present invention is respectively provided with image resolution ratio and is、、、, concentrate each point for being taken as parameter c in M by previous step, a large amount of different Julia collection figures can be generated
Picture.
3. a kind of improvement fractal image Underwater Image compression algorithm based on dictionary according to claim 1, its feature exists
In：The generation of dictionary and extension packets contain following steps in step 2：
（1）Segmentation Julia collection image generates various sizes of dictionary；
（2）Expand dictionary；
（3）Reject the redundancy in dictionary.
4. a kind of improvement fractal image Underwater Image compression algorithm based on dictionary according to claim 1, its feature exists
In：Dictionary classification considers following two factors for three kinds of various sizes of dictionaries in step 3, and following strategy is respectively adopted：
（1）Dictionary is classified
1. pressImage block each pixel is numbered, and sees Figure of description 3；
2. formula is pressedCalculate the characteristic value of each block；
（5-1）
In formula,；For numbering is in figurePoint pixel value；It is the pixel average of the block.
3. different classes are set according to characteristic value.Operation rule is 2. walked according to,The characteristic value of image block does not repeat to amount to
511（In the absence of the two-value block of full 0）, each characteristic value correspondenceA class in dictionary；
4. for each block in dictionary, according to its characteristic value, as corresponding to following rules decides whether to put it into it
Class（Remember that such is）In：
If a.It is empty set, is then directly placed into, otherwise, goes to b；
B. calculate the block withIn already present piece of PSNR（Y-PSNR）If result is respectively less than certain predetermined threshold value（Herein
Set）, andMiddle piece of quantity is not up to N, then be put into, and otherwise gives up；
The screening of this step is carried out, the block similarity included in dictionary can be avoided too high, in each class of increase between block and block
Otherness, is conducive to improving matching precision when encoding；
5. all pieces in dictionary are carried out with 4. step operation, dictionary classification is completed；
（2）Dictionary is classified
1. will by bilinearity difference approachThe size of image block is reduced into 3 × 3 by 4 × 4 in dictionary；
2. pressDictionary sorting technique is classified；
（3）Dictionary is classified
1. will by bilinearity difference approachIn dictionary the size of image block byIt is reduced into, it is rightImage block each pixel
Point is numbered, and sees Figure of description 4；
2. formula is pressed（5-2）Calculate the characteristic value of each block。
（5-2）
In formula,；；For numbering is in Fig. 5 .4Point pixel value；
It is the pixel average of the block；
3. operation rule is 2. walked according to,The characteristic value of image block does not repeat to amount to 8192, each characteristic value correspondenceA class in dictionary；
4. to each block in dictionary according toIn dictionary classification the 4., 5. step operation, complete screening and classify.
5. a kind of improvement fractal image Underwater Image compression algorithm based on dictionary according to claim 1, its feature exists
In：Comprised the steps of in step 4 coding：
（1）It is loaded into the dictionary according to the generation of step one, two and three；
（2）R block threshold values are set, match error threshold；
（3）Range block segmentation is carried out to image；
（4）Code book when selecting the dictionary of corresponding size as matching search according to range block cut size；
（5）For each R block, its variance is calculated, matching is scanned for as follows, obtain fractal code（Fractal code includes
Position and contrast factor of the best matching blocks in dictionaryWith grayscale shift amount）：
If 1., then it is smooth block to regard the R blocks, and any block in optional dictionary is best matching blocks, therefore without search
Fractal code can directly be exported；
If 2.Formula is then pressed according to R block sizesOr
Calculate the characteristic value of the R blocks, the class corresponding to it is selected in dictionary,In scan for matching, complete coding；
（6）Step is performed to all R blocks（5）, quantifying fractal code, output forms coding file.
6. a kind of improvement fractal image Underwater Image compression algorithm based on dictionary according to claim 1, its feature exists
In：Comprised the steps of in step 5 decoding：
（1）It is loaded into the dictionary by above-mentioned steps generation；
（2）By coding file read position in dictionary of R blocks, each corresponding best matching blocks of R blocks, contrast factor,
The information such as grayscale shift amount；
（3）One and an equal amount of image X of original image is defined, for preserving decoding image；
（4）For each the R block in X, its best matching blocks D, Ran Houyou are found in dictionary according to its fractal codeRecover；
（5）After all R blocks recover, composition decoding image is pieced together.
7. a kind of improvement fractal image Underwater Image compression algorithm based on dictionary according to claim 3, its feature exists
In：Segmentation Julia collection image generates various sizes of dictionary, the dictionary comprising following size：
1. dictionary is generated
A. will、Julia collection image segmentations areSub-block；
B. willJ collection image segmentations are、、Sub-block；WillJ collection image segmentations are、、、Sub-block, to more than generation sub-block, use the average method of pixel by its size change over for；
2. dictionary is generated
A. will、J collection image segmentations areSub-block；
B. willJ collection image segmentations are、, 21 × 21 sub-blocks；WillJ collection image segmentations are、12
×12、、, 42 × 42 sub-blocks, to more than generation sub-block, use the average method of pixel by its size change over for；
3. dictionary is generated
A. will、J collection image segmentation is 4 × 4 sub-blocks；
B. willJ collection image segmentations are、、Sub-block；WillJ collection image segmentations are、、、Sub-block, the sub-block of generation to more than, it is 4 × 4 by its size change over to use the average method of pixel.
8. a kind of improvement fractal image Underwater Image compression algorithm based on dictionary according to claim 3, its feature exists
In：
Expand dictionary and use following three kinds of modes：
1. prime number is multiplied：Each sub-block in dictionary is multiplied by different prime numbers（After can avoiding being multiplied with certain number from prime number
Block with other number be multiplied after block composition relation）, then to 256 modulus, new image block being generated, finally will newly scheme
As block is added in dictionary；
2. Gaussian noise is added.In view of Underwater Image generally existing noise, and the block after J collection image segmentations is noiseless
Image block, in order that dictionary is preferably matched with Underwater Image block, the present invention makes the block picture in dictionary by adding Gaussian noise
Element change has more randomness and diversity, to meet Underwater Image coding needs；
3. equilong transformation is carried out.To reduce matching error, with reference to basic fractal image process, the dictionary to generating carries out 8 kinds etc.
Away from conversion, expand dictionary capacity.
9. a kind of improvement fractal image Underwater Image compression algorithm based on dictionary according to claim 3, its feature exists
In：
Reject dictionary redundancy packet and contain following two steps：
1. the constant value block in dictionary is removed（There is a pixel value identical block）.Such piece can only be with when fractal image is matched
Constant value R Block- matchings, and constant value R blocks then can be with any Block- matching（Only need to be by contrast factorIt is set to 0）, therefore such piece not
Need to stay in dictionary；
2. the block with linear relationship in dictionary is removed.From fractal image matching process, with linear relationship
Block, matching effect is the same, only exists the difference in match parameter, therefore need to only retain one.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710030003.3A CN106815877A (en) | 2017-01-17 | 2017-01-17 | A kind of improvement fractal image Underwater Image compression algorithm based on dictionary |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710030003.3A CN106815877A (en) | 2017-01-17 | 2017-01-17 | A kind of improvement fractal image Underwater Image compression algorithm based on dictionary |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN106815877A true CN106815877A (en) | 2017-06-09 |
Family
ID=59111141
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710030003.3A Withdrawn CN106815877A (en) | 2017-01-17 | 2017-01-17 | A kind of improvement fractal image Underwater Image compression algorithm based on dictionary |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106815877A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108388670A (en) * | 2018-03-20 | 2018-08-10 | 华北电力大学（保定） | Quaternary tree based on local codebook divides the image search method of shape |
-
2017
- 2017-01-17 CN CN201710030003.3A patent/CN106815877A/en not_active Withdrawn
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108388670A (en) * | 2018-03-20 | 2018-08-10 | 华北电力大学（保定） | Quaternary tree based on local codebook divides the image search method of shape |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108335335B (en) | A kind of point cloud genera compression method based on enhancing figure transformation | |
| US6438268B1 (en) | Vector quantization codebook generation method | |
| CN111462261B (en) | Fast CU partitioning and intra-frame decision method for H.266/VVC | |
| CN111368662A (en) | Method, device, storage medium and equipment for editing attribute of face image | |
| KR20120095869A (en) | Method and apparatus for encoding a mesh model, encoded mesh model, and method and apparatus for decoding a mesh model | |
| CN1630876A (en) | Space-optimized texture maps | |
| CN112257766A (en) | Shadow recognition detection method under natural scene based on frequency domain filtering processing | |
| CN114493991A (en) | Attention circulation countermeasure network-based style migration system, method and device | |
| CN115829876A (en) | Real degraded image blind restoration method based on cross attention mechanism | |
| Constantinescu et al. | Improved techniques for single-pass adaptive vector quantization | |
| Wang et al. | Generative adversarial network based on resnet for conditional image restoration | |
| Ma et al. | Application of adaptive constructive neural networks to image compression | |
| CN113837191B (en) | Cross-star remote sensing image semantic segmentation method based on bidirectional unsupervised domain adaptive fusion | |
| Qu et al. | Low illumination enhancement for object detection in self-driving | |
| CN106815877A (en) | A kind of improvement fractal image Underwater Image compression algorithm based on dictionary | |
| Yan et al. | PlenVDB: Memory Efficient VDB-Based Radiance Fields for Fast Training and Rendering | |
| CN117132472A (en) | Forward-backward separable self-attention-based image super-resolution reconstruction method | |
| CN104299256A (en) | Almost-lossless compression domain volume rendering method for three-dimensional volume data | |
| CN115861779A (en) | Unbiased scene graph generation method based on effective feature representation | |
| CN113643400B (en) | Image generation method | |
| CN103297766B (en) | The compression method of vertex data and device in a kind of 3 d image data | |
| CN116977455A (en) | Face sketch image generation system and method based on deep two-way learning | |
| Choi et al. | Fast super-resolution algorithm using ELBP classifier | |
| Xu et al. | Residual blocks PointNet: A novel faster PointNet framework for segmentation and estimated pose | |
| Wang et al. | Real-time and accurate face detection networks based on deep learning |
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 | ||
| WW01 | Invention patent application withdrawn after publication | ||
| WW01 | Invention patent application withdrawn after publication |
Application publication date: 20170609 |