CN105141970B - A kind of texture image compression method based on three-dimensional model geometric information - Google Patents
A kind of texture image compression method based on three-dimensional model geometric information Download PDFInfo
- Publication number
- CN105141970B CN105141970B CN201510390491.XA CN201510390491A CN105141970B CN 105141970 B CN105141970 B CN 105141970B CN 201510390491 A CN201510390491 A CN 201510390491A CN 105141970 B CN105141970 B CN 105141970B
- Authority
- CN
- China
- Prior art keywords
- texture
- image
- point
- interest
- area
- 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.)
- Expired - Fee Related
Links
Abstract
It is furtherd investigate aiming at the problem that how to seek human eye area-of-interest on texture image in traditional limitation and compression process based on machine vision method product defects detection applicable object, the invention proposes a kind of texture image compression methods based on threedimensional model surface information, the human eye area-of-interest on texture image is sought using model meshes data, according to the Visual exhibition form of stage design of texture, the three-dimensional feature point and its mapping point on texture space that represent threedimensional model multiresolution gridding rear surface detailed information again are extracted as texture characteristic points, according to the continuity of image space, Type of Collective has been carried out to the characteristic point in image using K-means clustering algorithm, it achieves the area-of-interest of texture image and improves the differentiation precision of area-of-interest and background.Method of the invention is verified by building the EZW encoding and decoding experimental system based on ROI, obtains relatively good experiment effect.
Description
Technical field
The present invention relates to field of image processings, more particularly to a kind of texture image compression method.
Background technique
As Compute Graphics Theory basis and related software and hardware technology are constantly mature, dimensional Modeling Technology is gradually
Internet application has been incorporated, important application is achieved in web3D e-commerce, so that user in a network environment can also be with
Obtain the experience of similar real world.However, high-precision three-dimensional scanning and reconstruction technique are meeting user to virtual scene and object
While needs of the body Model sense of reality, also results in model data and data texturing amount is very huge.These data not only account for
According to a large amount of memory space, the speed of model transmission and load is also reduced, the experience of the network user is seriously affected, also restricts
The development of Web3D.Researcher by geological information and texture information compression to reduce data volume, by the development of many years,
Geological information compression aspect has been achieved for more research achievement, and the compression of threedimensional model texture image is still graphics neck
One of the hot issue of domain concern.According to the developing stage in Compression Study direction, texture compression algorithm can be divided into two classes: 1. bases
In the texture image compression algorithm of picture characteristics；2. the texture compression algorithm based on map feature.
In conclusion the texture image compression algorithm focus based on picture characteristics be the content of image, data type,
The various two dimensional characters such as frequency domain characteristic.And it is useless use data texturing by the generating process of three-dimensional space to two-dimensional space or
The various characteristics being mapped to by two-dimensional space during the textures of three-dimensional space.Texture compression algorithm based on map feature utilizes
Texture pixel completes some special operations to the characteristic in the mapping process of model surface, to the data in texture space.
Compression to grid data and texture image are generally divided into the research that 3 D graphic data carries out compressed encoding
The two relatively independent parts are compressed, few people's concern is associated both data using the associate feature of the two
Compressed encoding and compression.The present invention studies the cost for reducing texture regional area resolution ratio using grid geological information, utilizes
The geological information of grid determines the important area of texture image, realizes the preferential compression and biography of texture image regions of interest data
It is defeated.
Summary of the invention
In order to solve the problems in the prior art, for traditional based on applicable pair of the detection of machine vision method product defects
The problem of how seeking human eye area-of-interest on texture image in the limitation and compression process of elephant is furtherd investigate, this
Invention proposes a kind of texture graphics compression method based on three-dimensional model geometric information, and this method achieves the sense of texture image
Interest region and the differentiation precision for improving area-of-interest and background can determine texture image using the geological information of grid
Important area, realize texture image regions of interest data it is preferential compression and transmit.
The invention is realized by the following technical scheme:
A kind of texture graphics compression method based on three-dimensional model geometric information, it is characterised in that: the method is right first
Original figure is pre-processed, and determines interested region ROI, to generate ROI region module；Then ROI region template is utilized
Wavelet transformation is made to pretreated figure, is promoted or is reduced background coefficient to the corresponding wavelet conversion coefficient of ROI, connect down
Come to shape interested and location information coding transmission；Wherein it is determined that interested region ROI, to generate ROI region module tool
Body are as follows: seek the human eye area-of-interest on texture image using model meshes data and mentioned according to the Visual exhibition form of stage design of texture
It has taken and has represented the threedimensional model multiresolution three-dimensional feature point of gridding rear surface detailed information and its on texture space again
Mapping point is as texture characteristic points；According to the continuity of image space, using K-means clustering algorithm to the characteristic point in image
Type of Collective has been carried out, the area-of-interest of texture image is achieved and has improved the differentiation precision of area-of-interest and background.
As a further improvement of the present invention, 3-D graphic is handled with resolution method, model is divided into coarse base net
Lattice and many levels that different degrees of detailed information is added establish quad-tree structure for grid, facilitate and extract triangle gridding not
With the geological information under scale；The model meshes data use the half canonical weight grid based on multiple-limb of Igor Guskov
Change handles to obtain.
As a further improvement of the present invention, the background pixel in compression before removal textures in texture image.
As a further improvement of the present invention, the K-means clustering algorithm birdss of the same feather flock together to the characteristic point in image
It closes specifically: the set of characteristic points of texture are regarded as to the initial point set not clustered first, randomly selects a seed point and is used as to poly-
The central point of the class of conjunction；Then calculate each characteristic point to cluster centre distance, and cluster into the cluster nearest from the point
It goes；The coordinate average value of all the points in each cluster is then calculated, and using this average value as new cluster centre；Finally weigh
Each point is calculated again to the distance of cluster centre, and is referred in the class nearest from the point, until the result of algorithm restrains, note
The rectangular area of all characteristic points of class is completely surrounded in record.
As a further improvement of the present invention, described to be promoted or reduced background to the corresponding wavelet conversion coefficient of ROI
Coefficient specifically: the texture ROI region of extraction carries out the wavelet coefficient template of generation binaryzation after six layers of pyramid interative computation,
The wavelet coefficient for needing to retain includes the wavelet coefficient of area-of-interest and whole coefficients of low-frequency band in high frequency band；Using most
Big displacement method promotes or reduces the wavelet coefficient of processing.
Detailed description of the invention
Fig. 1 is the texture graphics compression method flow chart of the invention based on three-dimensional model geometric information；
Fig. 2 is the 1-4 sub-structure schematic diagram of triangular facet；
Fig. 3 is the quad-tree structure schematic diagram on side and vertex；
Fig. 4 is the difference value vector schematic diagram of vertex estimation position；
Fig. 5 is the feature point extraction result schematic diagram of texture image, wherein Fig. 5 (a) is the back of Buddha, Fig. 5 (b)
It is the head of Buddha, Fig. 5 (c) is the trunk decorative pattern of Rabbit, and Fig. 5 (d) is on the right side of the face of Rabbit, and Fig. 5 (e) is Bear
Body side, Fig. 5 (f) is the global texture of Bear, and Fig. 5 (g) is the face of Bear, the forelimb of Fig. 5 (h) Bear；
Fig. 6 is the texture image ROI region schematic diagram obtained by clustering algorithm, wherein Fig. 6 (a) is the pedestal of Bear,
Fig. 6 (b) is the head of Bear, and Fig. 6 (c) is the leg and foot of Bear, and Fig. 6 (d) is Bear overall situation textures；
Fig. 7 is the wavelet coefficient schematic diagram after maximum shift method is promoted, wherein Fig. 7 (a) is texture image, and Fig. 7 (b) is
Six layers of wavelet transformation black image, Fig. 7 (c) are wavelet coefficient climbing shutterings, and Fig. 7 (d) is the wavelet coefficient after being promoted.
The beneficial effects of the present invention are: the invention proposes a kind of, the texture image based on threedimensional model surface information compresses
Method is sought the human eye area-of-interest on texture image using model meshes data and is mentioned according to the Visual exhibition form of stage design of texture
It has taken and has represented the threedimensional model multiresolution three-dimensional feature point of gridding rear surface detailed information and its on texture space again
Mapping point is as texture characteristic points, according to the continuity of image space, using K-means clustering algorithm to the characteristic point in image
Type of Collective has been carried out, the area-of-interest of texture image is achieved and has improved the differentiation precision of area-of-interest and background.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that the specific embodiments described herein are merely illustrative of the present invention, and
It is not used in the restriction present invention.
Threedimensional model described in the present invention can be obtained from the three-dimensional modeling method based on image, can also be from energy
It enough generates and is obtained in the three-dimensional scanning device of texture.When carrying out texture image compression, it is necessary to have threedimensional model, texture image and line
Manage coordinate relationship.
It is attached it is shown in FIG. 1 be the texture graphics compression method process of the invention based on three-dimensional model geometric information, first
Original figure is pre-processed, then determines interested region ROI, to generate ROI region module, utilizes ROI region mould
Version makees wavelet transformation to pretreated figure, is promoted or is reduced background coefficient to the corresponding wavelet conversion coefficient of ROI, connect
Get off to shape interested and location information coding transmission.After the quantization encoding data for receiving transmission, to shape interested and
Location information decoding, then moves down the corresponding coefficient of area-of-interest, rebuilds figure after carrying out wavelet inverse transformation, preprocessing transformation
Picture.
Texture graphics compression method based on three-dimensional model geometric information of the invention seeks line using model meshes data
Human eye area-of-interest on reason image is extracted according to the Visual exhibition form of stage design of texture and represents threedimensional model multiresolution weight
The three-dimensional feature point of gridding rear surface detailed information and its mapping point on texture space are as texture characteristic points；According to figure
The continuity of image space has carried out Type of Collective to the characteristic point in image using K-means clustering algorithm, has achieved texture image
Area-of-interest and improve the differentiation precision of area-of-interest and background.
The state that texture image is presented on human eye is then different with normal image.On the one hand, the back before textures in texture image
Scene element, is absolutely not mapped to model surface, human eye when this partial information is for observing and nursing completely can not after textures
See, it may be considered that removed in compression.On the other hand, human eye focuses mostly on to the area-of-interest of model abundant in surface details
Region, in network transmission, this partial information should be preferentially passed to client.So in the compression process of texture image,
It should fully consider the prioritised transmission of area-of-interest.
It is often the inverse process of texture mapping by the transformation of texture coordinate system to device coordinate system in texture mapping.It gives
Determine space curved surface F ∈ R2, for any point (x, y, z) in F, its corresponding points in parameter field is found by texture mapping Φ
(u, v), it may be assumed that
Texture mapping function phi meets the condition for making input set and output set injection relationship each other.All Model Spaces
Between put coordinate (x, y, z) and its be all known part in three-dimensional modeling data in the respective coordinates (u, v) of texture space, therefore
The feature vertex in three-dimensional model gridding vertex might as well be first extracted, is then mapped that in texture coordinate system, you can get it line
Manage the characteristic point on image in area-of-interest.These characteristic points are centainly distributed in the visibility areas of texture after textures, and
The part of model surface geological information complexity interested to human eye is concentrated on, finally by the continuity of texture image content,
The area-of-interest of texture image can be solved.
The network topology information and its complexity of common threedimensional model, are unfavorable for seeking threedimensional model triangle gridding apex region
Characteristic point.3-D graphic is handled with resolution method, model is divided into coarse base net lattice and different degrees of details letter is added
The many levels of breath can establish quad-tree structure for grid, facilitate the geological information for extracting triangle gridding under different scale.
The present invention uses the gridding processing again of half canonical based on multiple-limb of Igor Guskov.Newly-generated grid has half canonical
Change topological structure, in addition to the out-degree (quantity on the side adjacent with vertex) on base meshing vertex is not 6, derives from the vertex in sub- face
The out-degree for being all 6 with numerical value.As shown in Fig. 2, the face of half regular net has natural 1-4 sub-structure, Ke Yijian
Vertical 1-4 spanning tree.Moreover, while and while on point similarly there is tree structure, attached drawing 3 illustrates this structure.
The Loop mode (Modified Loop) of application enhancements, is carried out in advance with vertex of the formula (2) to high-resolution level
Survey operation.As shown in Fig. 4, first with the position on vertex in higher level in the vertex estimation resolution ratio of low resolution level,
The position on high-rise vertex is replaced with the difference d of predicted position and physical location again.It should top down be handled in this way, then
Half regular net can turn to base net lattice vertex set M0With the offset on the vertex and its own predicted value of multilayer Progressive Mesh
Set di(i=1,2 ..., n), wherein n is the level number that base net lattice refine downwards.This representation method to grid to the greatest extent may be used
The low topology information for eliminating grid of energy, and remain the geological information of model surface.
When observing object, human eye is limited the resolution ratio of object detail, and attention rate may concentrate on some point
Resolution level, the information higher than the level are easily obtained and differentiate, and the details lower than the level will be difficult to be identified and pay close attention to.Institute
With when extracting the offset under grid different resolution, processing can meet the observation demand of human eye to a certain level.
Here, according to d1→…→dn-1→dnSequence, the offset wavelet coefficient set of the level of a certain suitable demand is done following
Operation: first to current collection di(i=1,2 ..., n, n are the offset sum of current level grid), if threshold valueSuccessively by diEach of wavelet coefficientAnd threshold epsiloniCompare, if wavelet coefficient is greater than
Threshold value then selects the characteristic point of three-dimensional point belonging to the wavelet coefficient alternately to save.Then it detects under current resolution
All qualified characteristic point numbers in set, if number is greater than preset value, (present invention considers that precision and operand will be standby
It selects characteristic point number to be set as 2000), is then twice current threshold value increasing, alternative features point is sought in repetition.It is total until point
Number is not more than preset value, finally saves these points as characteristic point.All preservation characteristic points are finally traversed, to each feature
Point vm, the smallest point v of its space length is found from the untreated original mesh of modelk, according to being each in original mesh
The mapping position in texture space that vertex provides, obtains corresponding pixel points T of this on two dimensional image spacevk, TvkAs
Required texture image characteristic point.
Aforesaid operations are applied on the texture image of several groups of threedimensional models, the series of results in attached drawing 5 can be obtained.Scheming
In it can clearly be observed that obtained texture characteristic points to be accurately distributed in the complex-shaped part of model geometric corresponding
Texture region, such as Rabbit in the head of Buddha in the fold of clothing in attached drawing 5 (a), attached drawing 5 (b), attached drawing 5 (c) and (d)
Face and decorative pattern with it and Fig. 5 (e), (g), (f), the regions such as the face of Bear and four limbs in (h).
Mapping by the grid search-engine point extraction of front and characteristic point to UV coordinate system, can obtain in texture image
The distribution of characteristic point.If it is possible to determine some regions, these regions are allowed completely to include characteristic point, so that it may obtain institute
The region of interest ROI needed.Positioned at the characteristic point of the same area, texture coordinate each other should be approached.It therefore can be with
The 2 d texture coordinate of picture feature point is analyzed as research object, according to the degree of closeness of coordinate to different characteristic point
Classify.The process that the set of physics or abstract object is divided into the multiple classes being made of similar object is referred to as and is clustered.
The ROI region formed by being clustered to image characteristic point can further the coefficient to image wavelet transform domain carry out
Processing.K-means clustering algorithm is suitable for solving the lesser two-dimensional problem of object scale just.
The set of characteristic points of texture are regarded as to the initial point set not clustered first, randomly selects n seed point and is used as to poly-
The central point of the class of conjunction；Then calculate each characteristic point to cluster centre distance, and cluster into the cluster nearest from the point
It goes；The coordinate average value of all the points in each cluster is then calculated, and using this average value as new cluster centre；Finally weigh
Each point is calculated again to the distance of cluster centre, and is referred in the class nearest from the point, until the result of algorithm restrains, note
The rectangular area of all characteristic points of class, i.e. Rectangular Bounding Volume are completely surrounded in record.As shown in attached drawing 6 (a)-(d), these in figure
The rectangular area that red edge comprising characteristic point surrounds is regarded as the area-of-interest of human eye on image.
The wavelet decomposition that step transverse direction is carried out to original image, with the mean value of low-frequency band storage original image adjacent pixel, high frequency
With reducing by half for storage Difference of Adjacent Pixels.Primary same operation is longitudinally being done again, then image completes one layer of wavelet decomposition, institute
There is low frequency coefficient to be all present in image upper left corner area.If continuing Haar wavelet transformation to image, only to original image
Low frequency part operation.Image contains the institute of the image pyramid of piece image different resolution after wavelet transformation
There is content.Wavelet transform function can't compress the data volume of original image, the number of pixels and original image of transformed image
Number of pixels it is equal.Original image needs realizing priority encoding to area-of-interest, it is necessary first to be promoted after transformation
Wavelet coefficient inside region；Also to reduce the coefficient of ROI region accordingly in decoding.The present invention uses maximum shift method,
Shift factor s meets s >=max (Mb), wherein MbIt is the maximum value of wavelet coefficient amplitude bit plane, s is displacement coefficient.It is flat through position
The minimum radius for the area-of-interest wavelet coefficient that face is promoted is greater than the maximum amplitude of background wavelet coefficient.With amplitude 2sFor boundary
Limit, it is background wavelet coefficient that amplitude, which is less than the value, and it is area-of-interest wavelet coefficient that amplitude, which is greater than the value,.So in attached drawing 7 (b)
Middle background color is black.After determining ROI region wavelet coefficient, these coefficients are moved to right s, ROI region coefficient can be restored.
The wavelet coefficient of generation binaryzation after six layers of pyramid interative computation is carried out to the texture ROI region that the present invention extracts
Template.Template size is identical with the texture image of Fig. 7 (a), and the white area of template corresponds in attached drawing 7 (c) needs in all frequency bands
Region where the wavelet coefficient to be retained.The wavelet coefficient for needing to retain includes the wavelet coefficient of area-of-interest in high frequency band
And whole coefficients of low-frequency band.For the promotion of low-frequency band background coefficient in attached drawing 7 (d), it will image is allowed to obtain when rebuilding
To preferable whole display effect.
Embedded Zerotree Wavelet Coding (EZW) takes full advantage of the statistical property of Wavelet image pyramidal morphology data, changes
The method for organizing of wavelet coefficient has been apt to it, ensure that important wavelet coefficient preferentially indicates and encode, and reduces to insignificant coefficient
Bit arithmetic number, and effectively raise reflection image reconstruction quality Y-PSNR.EZW selects initial threshold first
(threshold value that first scan obtains)Wherein | cI, j| be L grades of small echos decomposition coefficient, | cI, j| it is
cI, jAbsolute value, the threshold value scanned every time later is the half of previous scan threshold value:(i=1,2 ..., L), so
After be major-minor scanning, record sign bit P, N and zerotree root T and isolated point zero point Z of coefficient, and quantify to wavelet coefficient,
Final output encoded signal.
The wavelet coefficient application Embedded zero-tree wavelet compressed encoding that processing is promoted by maximum shift method compiles EZW
Code main scanning and the output of sub-scanning are as a result, can be by huffman coding second compression again, the binary system text that is finally encoded into
Part, the as compressed file of texture image.When network transmission, decoder can at any time be carried out image with the wavelet coefficient received
It rebuilds, decoding operate is considered as the inverse process of encoding operation.It is worth noting that, for by the small echo inside lifting region
Coefficient will also reduce the coefficient of ROI region in decoding accordingly.
Method of the invention is verified by building the EZW encoding and decoding experimental system based on ROI, is obtained relatively good
Experiment effect.To the threedimensional models such as Buddha, Rabbit, Bear by maximum shift method processing texture image wavelet coefficient into
The embedded compressed encoding of row and entropy coding, the texture dimensions of experimental model are entirely 4096 × 4096 pixels, uncompressed texture maps
The size of picture is respectively as follows: 33.4MB, 33.2MB and 33.1MB.It is extracted using area-of-interest exacting method proposed by the present invention special
Encoding region of interest processing is simultaneously done with Embedded zero-tree wavelet compression algorithm in sign region, obtains compressed file and is respectively as follows:
1.21MB, 1.09MB, 1.15MB compare the texture image file of original, uncompressed, compression ratio 3.6%, 3.3%, 3.5%.
The coefficient (ROI region coefficient) in compressed file by displacement processing is preferentially decoded and image in decoding end
Rebuild: for the texture image of Buddha model, background area temporarily only delivers low-frequency band coefficient to decoding end, lacks high frequency
Band wavelet coefficient, causes its resolution ratio to substantially reduce；And ROI region is all passed due to the coefficient that each frequency band is promoted by displacement
It is sent to decoding end, remains clarity well.
I.e. using a part of data seldom in compressed file (for Buddha compressed file be 32.0% data)
Preferable texture reconstruction effect and textures display effect can be obtained.Finally, compressed file is all decoded together with remainder.By
The precision of wavelet coefficient fractional digit is only had lost when carrying out Haar wavelet transformation to original texture image, and describes coefficient width
The integer part of value does not change, therefore can accurately be restored by these coefficients to image.
For experimental model, by transmitting 30%-50% file with high priority, it can preferably reappear former mould
The form of type, to the complicated part of model surface, the clarity of texture is higher.Continue to transmit remaining file and be pasted with file is rebuild
Figure, the display effect and original texture image that highest resolution can be obtained have no difference to the textures effect of grid model.
The above content is a further detailed description of the present invention in conjunction with specific preferred embodiments, and it cannot be said that
Specific implementation of the invention is only limited to these instructions.For those of ordinary skill in the art to which the present invention belongs, exist
Under the premise of not departing from present inventive concept, a number of simple deductions or replacements can also be made, all shall be regarded as belonging to of the invention
Protection scope.
Claims (5)
1. a kind of texture image compression method based on three-dimensional model geometric information, it is characterised in that: the method utilizes grid
Geological information determine the important area of texture image, realize the preferential compression of texture image area-of-interest and transmit；It is described
Method first pre-processes original figure, determines interested region ROI, to generate ROI region module；Then it utilizes
ROI region template makees wavelet transformation to pretreated figure, is promoted or reduced back to the corresponding wavelet conversion coefficient of ROI
Scape coefficient, next to shape interested and location information coding transmission；Wherein,
Carrying out pretreatment to original figure includes: to handle 3-D graphic with resolution method, and model is divided into coarse base net lattice
With many levels that different degrees of detailed information is added, quad-tree structure is established for grid, facilitates and extracts triangle gridding in difference
Geological information under scale；
Interested region ROI is determined, to generate ROI region module specifically: seek texture image using model meshes data
On human eye area-of-interest be extracted according to the Visual exhibition form of stage design of texture and represent the gridding again of threedimensional model multiresolution
The three-dimensional feature point of rear surface detailed information and its mapping point on texture space are as texture characteristic points；According to image space
Continuity, Type of Collective has been carried out to the characteristic point in image using K-means clustering algorithm, the sense for achieving texture image is emerging
Interesting region and the differentiation precision for improving area-of-interest and background.
2. according to the method described in claim 1, it is characterized by: the model meshes data use the base of Igor Guskov
In half canonical of multiple-limb, gridding handles to obtain again.
3. according to the method described in claim 1, it is characterized by: removing the background picture before textures in texture image in compression
Element.
4. according to the method described in claim 1, it is characterized by: the K-means clustering algorithm is to the characteristic point in image
Type of Collective is carried out specifically: the set of characteristic points of texture are regarded as to the initial point set not clustered first, randomly select n seed
Central point of the point as class to be polymerized；Then each characteristic point is calculated to the distance of cluster centre, and is clustered to most from the point
In close cluster；The coordinate average value of all the points in each cluster is then calculated, and using this average value as new cluster
Center；Each point is finally computed repeatedly to the distance of cluster centre, and is referred in the class nearest from the point, until algorithm
As a result it restrains, records the rectangular area for completely surrounding all characteristic points of class.
5. according to the method described in claim 1, it is characterized by: described promote the corresponding wavelet conversion coefficient of ROI
Or reduce background coefficient specifically: the texture ROI region of extraction generates the small of binaryzation after carrying out six layers of pyramid interative computation
Wave system digital-to-analogue plate, the wavelet coefficient for needing to retain include the whole of the wavelet coefficient of area-of-interest and low-frequency band in high frequency band
Coefficient；The wavelet coefficient of processing is promoted or reduced using maximum shift method.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510390491.XA CN105141970B (en) | 2015-07-03 | 2015-07-03 | A kind of texture image compression method based on three-dimensional model geometric information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201510390491.XA CN105141970B (en) | 2015-07-03 | 2015-07-03 | A kind of texture image compression method based on three-dimensional model geometric information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105141970A CN105141970A (en) | 2015-12-09 |
| CN105141970B true CN105141970B (en) | 2019-02-12 |
Family
ID=54727156
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201510390491.XA Expired - Fee Related CN105141970B (en) | 2015-07-03 | 2015-07-03 | A kind of texture image compression method based on three-dimensional model geometric information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105141970B (en) |
Families Citing this family (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10362290B2 (en) | 2015-02-17 | 2019-07-23 | Nextvr Inc. | Methods and apparatus for processing content based on viewing information and/or communicating content |
| US9729850B2 (en) | 2015-02-17 | 2017-08-08 | Nextvr Inc. | Methods and apparatus for receiving and/or using reduced resolution images |
| CN108604387A (en) * | 2016-02-11 | 2018-09-28 | 3M创新有限公司 | Surface mesh based on group is rebuild |
| CN107093197B (en) * | 2017-04-17 | 2020-11-17 | 浙江理工大学 | Animation compression method based on local cylindrical coordinates |
| CN107067426B (en) * | 2017-04-17 | 2018-02-06 | 东华理工大学 | A kind of image texture characteristic extracting method based on nested three-legged structure |
| CN107515227A (en) * | 2017-07-27 | 2017-12-26 | 武汉精测电子技术股份有限公司 | Based on the lossless method for compressing image of defect and defects of display panel detecting system |
| US11611774B2 (en) * | 2018-01-17 | 2023-03-21 | Sony Corporation | Image processing apparatus and image processing method for 3D data compression |
| CN109410139A (en) * | 2018-10-22 | 2019-03-01 | 湖南六新智能科技有限公司 | Inside a kind of historical relic and surface disease digital assay appraisal procedure |
| CN109255342B (en) * | 2018-11-20 | 2020-07-10 | 武汉大学 | Image region-of-interest extraction method and system based on two-step clustering of eye movement trajectory data |
| CN113473138B (en) * | 2021-06-30 | 2024-04-05 | 杭州海康威视数字技术股份有限公司 | Video frame coding method, device, electronic equipment and storage medium |
| CN117218263B (en) * | 2023-09-12 | 2024-03-19 | 山东捷瑞信息技术产业研究院有限公司 | Texture lightweight optimization method and system based on three-dimensional engine |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1655191A (en) * | 2005-02-25 | 2005-08-17 | 浙江大学 | Programmable graphic hardware based rapid voxel method for polygonal grid model |
| CN101877148A (en) * | 2009-11-13 | 2010-11-03 | 北京航空航天大学 | Method for repairing three-dimensional grid model based on global structure |
| CN101916454A (en) * | 2010-04-08 | 2010-12-15 | 董洪伟 | Method for reconstructing high-resolution human face based on grid deformation and continuous optimization |
-
2015
- 2015-07-03 CN CN201510390491.XA patent/CN105141970B/en not_active Expired - Fee Related
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1655191A (en) * | 2005-02-25 | 2005-08-17 | 浙江大学 | Programmable graphic hardware based rapid voxel method for polygonal grid model |
| CN101877148A (en) * | 2009-11-13 | 2010-11-03 | 北京航空航天大学 | Method for repairing three-dimensional grid model based on global structure |
| CN101916454A (en) * | 2010-04-08 | 2010-12-15 | 董洪伟 | Method for reconstructing high-resolution human face based on grid deformation and continuous optimization |
Non-Patent Citations (2)
| Title |
|---|
| Manifold-based approach to semi-regular remeshing;Guskov I;《Graphical Models》;20071231;第69卷(第1期);全文 |
| 基于小波和轮廓波变换的感兴趣区域编码研究;晏秀梅;《中国优秀硕士学位论文全文数据库-信息科技辑》;20110615(第06期);正文第3-4章 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105141970A (en) | 2015-12-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105141970B (en) | A kind of texture image compression method based on three-dimensional model geometric information | |
| Guo et al. | Image dehazing transformer with transmission-aware 3d position embedding | |
| US10552989B2 (en) | Point cloud attribute compression method based on KD tree and optimized graph transformation | |
| CN112396703B (en) | Reconstruction method of single-image three-dimensional point cloud model | |
| WO2019153342A1 (en) | Enhanced graph transformation-based point cloud attribute compression method | |
| Davoine et al. | Fractal image compression based on Delaunay triangulation and vector quantization | |
| Zhu et al. | Hifa: High-fidelity text-to-3d with advanced diffusion guidance | |
| TWI502547B (en) | Method and apparatus for planetary scale object rendering | |
| CN101626509A (en) | Methods and devices for encoding and decoding three dimensional grids | |
| CN102129698A (en) | Image coding method based on region of interest | |
| CN103971329A (en) | Cellular nerve network with genetic algorithm (GACNN)-based multisource image fusion method | |
| Xu et al. | Cluster-based point cloud coding with normal weighted graph fourier transform | |
| CN101447084A (en) | Compressing encode/decode method for three-dimensional model texture based on Quardtree | |
| CN115471634A (en) | Modeling method and device for urban green plant twins | |
| Mao et al. | Generalization of 3D building texture using image compression and multiple representation data structure | |
| JP4398785B2 (en) | Multidimensional data encoding method, multidimensional data decoding method, texture image creation method, apparatus for realizing the methods, and program for realizing the methods | |
| Berjón et al. | Objective and subjective evaluation of static 3D mesh compression | |
| CN110322548A (en) | A kind of three-dimensional grid model generation method based on several picture parametrization | |
| CN112785684B (en) | Three-dimensional model reconstruction method based on local information weighting mechanism | |
| CN103024377A (en) | Synthetic aperture radar (SAR) image compression method based on target area extraction and direction wave | |
| He et al. | Point set surface compression based on shape pattern analysis | |
| Wang et al. | Progressive point set surface compression based on planar reflective symmetry analysis | |
| Song et al. | City-on-Web: Real-time Neural Rendering of Large-scale Scenes on the Web | |
| Friedman et al. | Automatic procedural modeling of tree structures in point clouds using wavelets | |
| Zhao et al. | A pencil drawing algorithm based on wavelet transform multiscale |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CF01 | Termination of patent right due to non-payment of annual fee |
Granted publication date: 20190212 Termination date: 20190703 |
|
| CF01 | Termination of patent right due to non-payment of annual fee |